Abhishek Kumar

Yannan Wang, Longli Yang, Zhen Liu et al.

Deploying lightweight Large Language Model (LLM) agents on edge servers can reduce latency and move agentic services closer to users, but resource-constrained edge models often struggle with long-horizon tasks that require persistent memory, subgoal tracking, and reflection. Fine-tuning edge models after deployment is costly and difficult to scale across heterogeneous nodes, while purely local memory leaves agents with isolated experience and growing prompt context. We propose \textsc{CoMIC}, a parameter-update-free cloud-edge framework for Collaborative Memory and Insights Circulation. \textsc{CoMIC} follows a \textit{Centralized Reflection, Decentralized Execution} design: edge agents execute locally using subgoal-oriented hierarchical memory and selective re-expansion of relevant histories, while a cloud-side LLM critic asynchronously evaluates completed trajectories, filters reusable experience, and aggregates cross-agent guidance keyed by semantic subgoal identifiers. Across five long-horizon agent tasks spanning symbolic planning and text interaction, \textsc{CoMIC} improves progress rate and action grounding for weak edge agents and yields task-dependent success-rate gains without updating model parameters.

Abhishek Kumar, Isha Motiyani, Tilak Kasturi et al.

Engineering diagrams pose a distinct challenge for vision-language models: unlike natural images or general documents, they encode information through dense spatial layouts, domain-specific symbols, and cross-references between visual callouts and structured parts tables. Despite their centrality to service, repair, and design workflows, there is no public benchmark for measuring VLM capabilities in this domain; existing datasets primarily focus on flowcharts, scientific figures, or business documents. To address this gap, we introduce Enginuity, the first open dataset and benchmark for evaluating VLMs on complex engineering diagrams. We define two tasks over a corpus of U.S. military service and repair manuals: structured parts-table extraction (Task 1) and free-form visual diagram question answering (VQA)(Task 2) for benchmarking. We evaluate four frontier VLMs (GPT-5.2 Chat, Claude Opus 4.7, Gemma 4, Qwen3-VL-32B-Instruct) under zero-shot and chain-of-thought prompting. On Task 1, models reach Recall@all of 0.61-0.87 but Token F1pen of only 0.03-0.18, exposing a systematic gap between part identification and description fidelity. Task 2 reveals a consistent factual-reasoning gap across all models. A supporting analysis shows that token-overlap metrics under-report model capability on technical descriptions by 2-6x relative to semantic similarity, motivating LLM-as-judge calibration for domain-specific evaluation. We release the dataset, annotations, evaluation harness, and per-sample model outputs to support a reproducible study of VLM capability on engineering content.

Mitchell Wortsman, Peter J. Liu, Lechao Xiao et al. · anthropic, deepmind

Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smaller scales. Although the causes of such instabilities are of scientific interest, the amount of resources required to reproduce them has made investigation difficult. In this work, we seek ways to reproduce and study training stability and instability at smaller scales. First, we focus on two sources of training instability described in previous work: the growth of logits in attention layers (Dehghani et al., 2023) and divergence of the output logits from the log probabilities (Chowdhery et al., 2022). By measuring the relationship between learning rate and loss across scales, we show that these instabilities also appear in small models when training at high learning rates, and that mitigations previously employed at large scales are equally effective in this regime. This prompts us to investigate the extent to which other known optimizer and model interventions influence the sensitivity of the final loss to changes in the learning rate. To this end, we study methods such as warm-up, weight decay, and the $μ$Param (Yang et al., 2022), and combine techniques to train small models that achieve similar losses across orders of magnitude of learning rate variation. Finally, to conclude our exploration we study two cases where instabilities can be predicted before they emerge by examining the scaling behavior of model activation and gradient norms.

Jiaheng Wei, Harikrishna Narasimhan, Ehsan Amid et al.

The generalization ability of machine learning models degrades significantly when the test distribution shifts away from the training distribution. We investigate the problem of training models that are robust to shifts caused by changes in the distribution of class-priors or group-priors. The presence of skewed training priors can often lead to the models overfitting to spurious features. Unlike existing methods, which optimize for either the worst or the average performance over classes or groups, our work is motivated by the need for finer control over the robustness properties of the model. We present an extremely lightweight post-hoc approach that performs scaling adjustments to predictions from a pre-trained model, with the goal of minimizing a distributionally robust loss around a chosen target distribution. These adjustments are computed by solving a constrained optimization problem on a validation set and applied to the model during test time. Our constrained optimization objective is inspired by a natural notion of robustness to controlled distribution shifts. Our method comes with provable guarantees and empirically makes a strong case for distributional robust post-hoc classifiers. An empirical implementation is available at https://github.com/weijiaheng/Drops.

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Tyler LaBonte, Vidya Muthukumar, Abhishek Kumar

Empirical risk minimization (ERM) of neural networks is prone to over-reliance on spurious correlations and poor generalization on minority groups. The recent deep feature reweighting (DFR) technique achieves state-of-the-art group robustness via simple last-layer retraining, but it requires held-out group and class annotations to construct a group-balanced reweighting dataset. In this work, we examine this impractical requirement and find that last-layer retraining can be surprisingly effective with no group annotations (other than for model selection) and only a handful of class annotations. We first show that last-layer retraining can greatly improve worst-group accuracy even when the reweighting dataset has only a small proportion of worst-group data. This implies a "free lunch" where holding out a subset of training data to retrain the last layer can substantially outperform ERM on the entire dataset with no additional data or annotations. To further improve group robustness, we introduce a lightweight method called selective last-layer finetuning (SELF), which constructs the reweighting dataset using misclassifications or disagreements. Our empirical and theoretical results present the first evidence that model disagreement upsamples worst-group data, enabling SELF to nearly match DFR on four well-established benchmarks across vision and language tasks with no group annotations and less than 3% of the held-out class annotations. Our code is available at https://github.com/tmlabonte/last-layer-retraining.

Henna Kokkonen, Lauri Lovén, Naser Hossein Motlagh et al.

Future AI applications require performance, reliability and privacy that the existing, cloud-dependant system architectures cannot provide. In this article, we study orchestration in the device-edge-cloud continuum, and focus on edge AI for resource orchestration. We claim that to support the constantly growing requirements of intelligent applications in the device-edge-cloud computing continuum, resource orchestration needs to embrace edge AI and emphasize local autonomy and intelligence. To justify the claim, we provide a general definition for continuum orchestration, and look at how current and emerging orchestration paradigms are suitable for the computing continuum. We describe certain major emerging research themes that may affect future orchestration, and provide an early vision of an orchestration paradigm that embraces those research themes. Finally, we survey current key edge AI methods and look at how they may contribute into fulfilling the vision of future continuum orchestration.

Tyler LaBonte, John C. Hill, Xinchen Zhang et al.

Modern machine learning models are prone to over-reliance on spurious correlations, which can often lead to poor performance on minority groups. In this paper, we identify surprising and nuanced behavior of finetuned models on worst-group accuracy via comprehensive experiments on four well-established benchmarks across vision and language tasks. We first show that the commonly used class-balancing techniques of mini-batch upsampling and loss upweighting can induce a decrease in worst-group accuracy (WGA) with training epochs, leading to performance no better than without class-balancing. While in some scenarios, removing data to create a class-balanced subset is more effective, we show this depends on group structure and propose a mixture method which can outperform both techniques. Next, we show that scaling pretrained models is generally beneficial for worst-group accuracy, but only in conjunction with appropriate class-balancing. Finally, we identify spectral imbalance in finetuning features as a potential source of group disparities -- minority group covariance matrices incur a larger spectral norm than majority groups once conditioned on the classes. Our results show more nuanced interactions of modern finetuned models with group robustness than was previously known. Our code is available at https://github.com/tmlabonte/revisiting-finetuning.

Burak Varici, Emre Acarturk, Karthikeyan Shanmugam et al.

This paper studies the causal representation learning problem when the latent causal variables are observed indirectly through an unknown linear transformation. The objectives are: (i) recovering the unknown linear transformation (up to scaling) and (ii) determining the directed acyclic graph (DAG) underlying the latent variables. Sufficient conditions for DAG recovery are established, and it is shown that a large class of non-linear models in the latent space (e.g., causal mechanisms parameterized by two-layer neural networks) satisfy these conditions. These sufficient conditions ensure that the effect of an intervention can be detected correctly from changes in the score. Capitalizing on this property, recovering a valid transformation is facilitated by the following key property: any valid transformation renders latent variables' score function to necessarily have the minimal variations across different interventional environments. This property is leveraged for perfect recovery of the latent DAG structure using only \emph{soft} interventions. For the special case of stochastic \emph{hard} interventions, with an additional hypothesis testing step, one can also uniquely recover the linear transformation up to scaling and a valid causal ordering.

Boning Li, Timofey Efimov, Abhishek Kumar et al.

Network digital twins (NDTs) facilitate the estimation of key performance indicators (KPIs) before physically implementing a network, thereby enabling efficient optimization of the network configuration. In this paper, we propose a learning-based NDT for network simulators. The proposed method offers a holistic representation of information flow in a wireless network by integrating node, edge, and path embeddings. Through this approach, the model is trained to map the network configuration to KPIs in a single forward pass. Hence, it offers a more efficient alternative to traditional simulation-based methods, thus allowing for rapid experimentation and optimization. Our proposed method has been extensively tested through comprehensive experimentation in various scenarios, including wired and wireless networks. Results show that it outperforms baseline learning models in terms of accuracy and robustness. Moreover, our approach achieves comparable performance to simulators but with significantly higher computational efficiency.

Kenneth V. Price, Abhishek Kumar, Ponnuthurai N Suganthan

Non-parametric tests can determine the better of two stochastic optimization algorithms when benchmarking results are ordinal, like the final fitness values of multiple trials. For many benchmarks, however, a trial can also terminate once it reaches a pre-specified target value. When only some trials reach the target value, two variables characterize a trial's outcome: the time it takes to reach the target value (or not) and its final fitness value. This paper describes a simple way to impose linear order on this two-variable trial data set so that traditional non-parametric methods can determine the better algorithm when neither dominates. We illustrate the method with the Mann-Whitney U-test. A simulation demonstrates that U-scores are much more effective than dominance when tasked with identifying the better of two algorithms. We test U-scores by having them determine the winners of the CEC 2022 Special Session and Competition on Real-Parameter Numerical Optimization.

Abhishek Kumar, Oladayo S. Ajani, Swagatam Das et al.

In machine learning and computer vision, mean shift (MS) qualifies as one of the most popular mode-seeking algorithms used for clustering and image segmentation. It iteratively moves each data point to the weighted mean of its neighborhood data points. The computational cost required to find the neighbors of each data point is quadratic to the number of data points. Consequently, the vanilla MS appears to be very slow for large-scale datasets. To address this issue, we propose a mode-seeking algorithm called GridShift, with significant speedup and principally based on MS. To accelerate, GridShift employs a grid-based approach for neighbor search, which is linear in the number of data points. In addition, GridShift moves the active grid cells (grid cells associated with at least one data point) in place of data points towards the higher density, a step that provides more speedup. The runtime of GridShift is linear in the number of active grid cells and exponential in the number of features. Therefore, it is ideal for large-scale low-dimensional applications such as object tracking and image segmentation. Through extensive experiments, we showcase the superior performance of GridShift compared to other MS-based as well as state-of-the-art algorithms in terms of accuracy and runtime on benchmark datasets for image segmentation. Finally, we provide a new object-tracking algorithm based on GridShift and show promising results for object tracking compared to CamShift and meanshift++.

Sayan Saha, Saptarshi Das, Anish Acharya et al.

The paper investigates nonlinear system identification using system output data at various linearized operating points. A feed-forward multi-layer Artificial Neural Network (ANN) based approach is used for this purpose and tested for two target applications i.e. nuclear reactor power level monitoring and an AC servo position control system. Various configurations of ANN using different activation functions, number of hidden layers and neurons in each layer are trained and tested to find out the best configuration. The training is carried out multiple times to check for consistency and the mean and standard deviation of the root mean square errors (RMSE) are reported for each configuration.

Jiaheng Wei, Zhaowei Zhu, Tianyi Luo et al.

The rawly collected training data often comes with separate noisy labels collected from multiple imperfect annotators (e.g., via crowdsourcing). A typical way of using these separate labels is to first aggregate them into one and apply standard training methods. The literature has also studied extensively on effective aggregation approaches. This paper revisits this choice and aims to provide an answer to the question of whether one should aggregate separate noisy labels into single ones or use them separately as given. We theoretically analyze the performance of both approaches under the empirical risk minimization framework for a number of popular loss functions, including the ones designed specifically for the problem of learning with noisy labels. Our theorems conclude that label separation is preferred over label aggregation when the noise rates are high, or the number of labelers/annotations is insufficient. Extensive empirical results validate our conclusions.

Sunny Sanyal, Atula Neerkaje, Jean Kaddour et al.

Training Large Language Models (LLMs) incurs significant cost; hence, any strategy that accelerates model convergence is helpful. In this paper, we investigate the ability of a simple idea checkpoint averaging along the trajectory of a training run to improve both convergence and generalization quite early on during training. Here we show that models trained with high learning rates observe higher gains due to checkpoint averaging. Furthermore, these gains are amplified when checkpoints are sampled with considerable spacing in training steps. Our training recipe outperforms conventional training and popular checkpoint averaging baselines such as exponential moving average (EMA) and stochastic moving average (SWA). We evaluate our training recipe by pre-training LLMs, where high learning rates are inherently preferred due to extremely large batch sizes. Specifically, we pre-trained nanoGPT-2 models of varying sizes, small (125M), medium (335M), and large (770M)on the OpenWebText dataset, comprised of 9B tokens. Additionally, we present results for publicly available Pythia LLMs, ranging from 1B to 12B, which were trained on the PILE-deduped dataset containing 207B tokens.

Lauri Lovén, Roberto Morabito, Abhishek Kumar et al.

The edge-cloud computing continuum demands self-management mechanisms that scale across autonomous administrative domains while honouring tenant- and operator-specified data sovereignty. We present Neural Pub/Sub, a federated-broker autonomic substrate whose self-organising behaviour emerges from market-based price signals rather than centralised control. Its MAPE-K control loop closes over per-broker health and load monitoring, marginal-cost clearing-price analysis, placement planning over a polymatroidal feasibility region, federated cross-domain dispatch, and shared peer subscription summaries with bounded-staleness price signals. The Plan step is anchored in a Walrasian convergence proposition: under gross-substitutes valuations on tree and series-parallel service-dependency DAGs, decentralised price-based allocation matches the welfare of a centralised oracle. We evaluate the substrate on a 4-VM, 4-domain, 48-worker federated edge-cloud testbed (single data centre, 50 ms emulated WAN) in a 1005-run campaign augmented by a fair-process-count sharded-oracle comparator. The federated market dominates a single-process oracle by 2-4% with 45 of 45 per-seed wins (sign-test p ~ 2.8e-14, Hodges-Lehmann median -39.6 ms); against a four-shard centralised orchestrator at equal process count the gap stays within +/-1.5% across all nine (pipeline, load) cells. Round-robin completion rate collapses 98.8% -> 22.4% -> 3.3% across arrival rates 5/10/15 pps while the market preserves completion; the advantage decomposes into three Walrasian properties (information completeness, admission control, price discovery). Federation withstands broker death and network partition (completion rate >= 98.7% across 75 cells), and sovereignty enforcement adds no measurable runtime overhead across 60 governance-grid runs. Heterogeneous-domain stressors and cross-site WAN deployment remain future work.

Lauri Lovén, Abhishek Kumar, Alexander Engelhardt et al.

Large language models (LLMs) can serve as the semantic-matching engine of a content-based publish/subscribe broker for agentic AI across the edge-cloud computing continuum, bridging the vocabulary and modality gaps that defeat keyword and embedding filters. Framed as offline multi-label retrieval over three public datasets spanning social-media, legal, and smart-home sensor domains (six LLMs, seven baselines), our central contribution is a two-crossover cost-accuracy characterisation: an analytical context-window crossover below which a CoverAndMerge compression pipeline reduces LLM invocations, and an empirical discrimination-capacity crossover above which matching accuracy collapses independently of context budget, by a model-dependent factor of parameter count and training generation. Two findings carry practical weight: above the discrimination crossover, compression cannot recover accuracy and only frontier-scale models clear large subscription sets; and there backend choice dominates configuration choice, so model selection, not pipeline tuning, is the primary operator lever. We accompany this with three composable algorithms and a per-cluster Quality-of-Experience framework for autonomic LLM-tier selection.

Carsten Maple, Abhishek Kumar, Riya Tapwal

Many jailbreak attack research papers report attack success rates for a limited number of parameter settings, even though there are many combinations of parameter settings that could be used. Further, when new jailbreak papers are released, they often benchmark results against single configurations of existing attacks. This position paper argues such practices are fundamentally insufficient for characterising the threat posed by parameterised jailbreak attacks, and comparing attacks. Most jailbreak attacks expose multiple internal parameters, system prompt templates, conversation rounds, cipher dispersion, teaching shots, and ASR varies substantially across these parameters. Reporting only the best-case configuration discards two pieces of information that defenders genuinely need: how typical that performance is across the variant space, and how much of the attack surface is missed by selecting a single variant. We propose two new measures for jailbreak attacks: the Variant Sensitivity Measure (VSM) and Union Coverage (UC). VSM quantifies how far the best reported ASR deviates from the mean ASR across the tested variant space, UC is the total fraction of prompts resulting in unsafe responses across all tested configurations. We empirically demonstrate the importance of these measures using two attack families across three open-source target models. For PAIR, the best template reaches 69% ASR on Mistral-7B and 75% on Qwen3-0.6B, while UC rises to 88% and 93%, respectively. For bijection on Mistral-7B, the best variant reaches 81% ASR, but the 36-variant union covers 100% of HarmBench-100 prompts. We argue that distributional reporting, publishing VSM alongside ASR and enumerating variant coverage as fully as compute allows, should become the new minimum standard for parameterised jailbreak evaluation.

Athul Shibu, Abhishek Kumar, Heechul Jung et al.

Convolutional Neural Networks (CNNs) have a large number of parameters and take significantly large hardware resources to compute, so edge devices struggle to run high-level networks. This paper proposes a novel method to reduce the parameters and FLOPs for computational efficiency in deep learning models. We introduce accuracy and efficiency coefficients to control the trade-off between the accuracy of the network and its computing efficiency. The proposed Rewarded meta-pruning algorithm trains a network to generate weights for a pruned model chosen based on the approximate parameters of the final model by controlling the interactions using a reward function. The reward function allows more control over the metrics of the final pruned model. Extensive experiments demonstrate superior performances of the proposed method over the state-of-the-art methods in pruning ResNet-50, MobileNetV1, and MobileNetV2 networks.

Satyawant Kumar, Abhishek Kumar, Dong-Gyu Lee

Remotely captured images possess an immense scale and object appearance variability due to the complex scene. It becomes challenging to capture the underlying attributes in the global and local context for their segmentation. Existing networks struggle to capture the inherent features due to the cluttered background. To address these issues, we propose a remote sensing image segmentation network, RemoteNet, for semantic segmentation of remote sensing images. We capture the global and local features by leveraging the benefits of the transformer and convolution mechanisms. RemoteNet is an encoder-decoder design that uses multi-scale features. We construct an attention map module to generate channel-wise attention scores for fusing these features. We construct a global-local transformer block (GLTB) in the decoder network to support learning robust representations during a decoding phase. Further, we designed a feature refinement module to refine the fused output of the shallow stage encoder feature and the deepest GLTB feature of the decoder. Experimental findings on the two public datasets show the effectiveness of the proposed RemoteNet.

Hye-Seong Hong, Abhishek Kumar, Dong-Gyu Lee

The generalization capability of unsupervised domain adaptation can mitigate the need for extensive pixel-level annotations to train semantic segmentation networks by training models on synthetic data as a source with computer-generated annotations. Entropy-based adversarial networks are proposed to improve source domain prediction; however, they disregard significant external information, such as edges, which have the potential to identify and distinguish various objects within an image accurately. To address this issue, we introduce a novel approach to domain adaptation, leveraging the synergy of internal and external information within entropy-based adversarial networks. In this approach, we enrich the discriminator network with edge-predicted probability values within this innovative framework to enhance the clarity of class boundaries. Furthermore, we devised a probability-sharing network that integrates diverse information for more effective segmentation. Incorporating object edges addresses a pivotal aspect of unsupervised domain adaptation that has frequently been neglected in the past -- the precise delineation of object boundaries. Conventional unsupervised domain adaptation methods usually center around aligning feature distributions and may not explicitly model object boundaries. Our approach effectively bridges this gap by offering clear guidance on object boundaries, thereby elevating the quality of domain adaptation. Our approach undergoes rigorous evaluation on the established unsupervised domain adaptation benchmarks, specifically in adapting SYNTHIA $\rightarrow$ Cityscapes and SYNTHIA $\rightarrow$ Mapillary. Experimental results show that the proposed model attains better performance than state-of-the-art methods. The superior performance across different unsupervised domain adaptation scenarios highlights the versatility and robustness of the proposed method.

Mahesh Bhosale, Abhishek Kumar, David Doermann

We propose a neural network architecture that learns body part appearances for soccer player re-identification. Our model consists of a two-stream network (one stream for appearance map extraction and the other for body part map extraction) and a bilinear-pooling layer that generates and spatially pools the body part map. Each local feature of the body part map is obtained by a bilinear mapping of the corresponding local appearance and body part descriptors. Our novel representation yields a robust image-matching feature map, which results from combining the local similarities of the relevant body parts with the weighted appearance similarity. Our model does not require any part annotation on the SoccerNet-V3 re-identification dataset to train the network. Instead, we use a sub-network of an existing pose estimation network (OpenPose) to initialize the part substream and then train the entire network to minimize the triplet loss. The appearance stream is pre-trained on the ImageNet dataset, and the part stream is trained from scratch for the SoccerNet-V3 dataset. We demonstrate the validity of our model by showing that it outperforms state-of-the-art models such as OsNet and InceptionNet.

Riya Tapwal, Abhishek Kumar, Carsten Maple

Large language models (LLMs) are increasingly used as automatic judges for summarization and dialogue evaluation. Prior work has documented biases such as position, verbosity, and style preferences, but largely focuses on outcomes, leaving judge explanations underexplored. We instead ask whether LLM judges are cue-invariant, i.e., whether their rankings and explanations remain stable when non-evidential cues are perturbed while holding the underlying texts fixed. We introduce a suite of cue interventions (Blind, Truth, Flip, Placebo, Reveal-After) and tie-aware metrics that quantify outcome anchoring and rationale anchoring, including label-aligned rhetoric and explanation drift, alongside consistency and stereotype-intrusion checks. We design anchoring attacks using verbosity and confidence cues, and compare two mitigations: structured chain-of-thought prompting and PROOF-BEFORE-PREFERENCE (evidence lock, score, rank). Using a new dataset of 1,000 summaries from traditional extractive models and LLMs, we find substantial cue-anchored rationalization under label and placebo perturbations, while PROOF-BEFORE-PREFERENCE markedly improves cue invariance over baselines.

Avi Singh, John D. Co-Reyes, Rishabh Agarwal et al. · anthropic, deepmind

Fine-tuning language models~(LMs) on human-generated data remains a prevalent practice. However, the performance of such models is often limited by the quantity and diversity of high-quality human data. In this paper, we explore whether we can go beyond human data on tasks where we have access to scalar feedback, for example, on math problems where one can verify correctness. To do so, we investigate a simple self-training method based on expectation-maximization, which we call ReST$^{EM}$, where we (1) generate samples from the model and filter them using binary feedback, (2) fine-tune the model on these samples, and (3) repeat this process a few times. Testing on advanced MATH reasoning and APPS coding benchmarks using PaLM-2 models, we find that ReST$^{EM}$ scales favorably with model size and significantly surpasses fine-tuning only on human data. Overall, our findings suggest self-training with feedback can substantially reduce dependence on human-generated data.

Yuhang Zhou, Zhuokai Zhao, Ke Li et al.

Feature selection is a crucial step in large-scale industrial machine learning systems, directly affecting model accuracy, efficiency, and maintainability. Traditional feature selection methods rely on labeled data and statistical heuristics, making them difficult to apply in production environments where labeled data are limited and multiple operational constraints must be satisfied. To address this, we propose Model Feature Agent (MoFA), a model-driven framework that performs sequential, reasoning-based feature selection using both semantic and quantitative feature information. MoFA incorporates feature definitions, importance scores, correlations, and metadata (e.g., feature groups or types) into structured prompts and selects features through interpretable, constraint-aware reasoning. We evaluate MoFA in three real-world industrial applications: (1) True Interest and Time-Worthiness Prediction, where it improves accuracy while reducing feature group complexity, (2) Value Model Enhancement, where it discovers high-order interaction terms that yield substantial engagement gains in online experiments, and (3) Notification Behavior Prediction, where it selects compact, high-value feature subsets that improve both model accuracy and inference efficiency. Together, these results demonstrate the practicality and effectiveness of LLM-based reasoning for feature selection in real production systems.

Riya Tapwal, Abhishek Kumar, Carsten Maple

Multi-agent LLM systems introduce a security risk in which sensitive information accessed by one agent can propagate through shared context and reappear in downstream outputs, even without explicit adversarial intent. We formalise this phenomenon as propagation amplification, where leakage risk increases across agent boundaries as sensitive content is repeatedly exposed to downstream generators. Existing defences, including prompt-based safeguards, static pattern matching, and LLM-as-judge filtering, are not designed for this setting: they either operate after generation, rely primarily on surface-form patterns, or add substantial latency without modelling the generation process itself. To resolve these issues, we propose PRISM, a real-time defence that treats credential leakage as a sequential risk accumulation problem during generation. At each decoding step, PRISM combines 16 signals spanning lexical, structural, information-theoretic, behavioural, and contextual features into a calibrated risk score, enabling per-token intervention through green, yellow, and red risk zones. Our central observation is that credential reproduction is often preceded by a measurable shift in generation dynamics, characterised by entropy collapse and increasing logit concentration. When combined with text-structural cues such as identifier-pattern detection, these temporal signals provide an early warning of leakage before a secret is fully reconstructed. Across a 2,000-task adversarial benchmark covering 13 attack categories and three pressure levels in a heterogeneous four-agent pipeline, PRISM achieves F1 = 0.832 with precision = 1.000 and recall = 0.712, while producing no observed leakage on our benchmark (0.0% task-level leak rate) and preserving output utility of 0.893. It substantially outperforms the strongest baseline, Span Tagger, which achieves F1 = 0.719 with a 15.0% task-level leak rate.

Abhishek Kumar, Dong-Gyu Lee

Contemporary deep clustering approaches often rely on either contrastive or non-contrastive techniques to acquire effective representations for clustering tasks. Contrastive methods leverage negative pairs to achieve homogenous representations but can introduce class collision issues, potentially compromising clustering performance. On the contrary, non-contrastive techniques prevent class collisions but may produce non-uniform representations that lead to clustering collapse. In this work, we propose a novel end-to-end deep clustering approach named PIPCDR, designed to harness the strengths of both approaches while mitigating their limitations. PIPCDR incorporates a positive instance proximity loss and a cluster dispersion regularizer. The positive instance proximity loss ensures alignment between augmented views of instances and their sampled neighbors, enhancing within-cluster compactness by selecting genuinely positive pairs within the embedding space. Meanwhile, the cluster dispersion regularizer maximizes inter-cluster distances while minimizing within-cluster compactness, promoting uniformity in the learned representations. PIPCDR excels in producing well-separated clusters, generating uniform representations, avoiding class collision issues, and enhancing within-cluster compactness. We extensively validate the effectiveness of PIPCDR within an end-to-end Majorize-Minimization framework, demonstrating its competitive performance on moderate-scale clustering benchmark datasets and establishing new state-of-the-art results on large-scale datasets.

Kevin Han, Yuhang Zhou, Mingze Gao et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has proven effective for enhancing the reasoning capabilities of Large Language Models (LLMs). However, dominant approaches like Group Relative Policy Optimization (GRPO) face critical stability challenges: they suffer from high estimator variance under computational constraints (small group sizes) and vanishing gradient signals in saturated failure regimes where all responses yield identical zero rewards. To address this, we propose Empirical Bayes Policy Optimization (EBPO), a novel framework that regularizes local group-based baselines by borrowing strength from the policy's accumulated global statistics. Instead of estimating baselines in isolation, EBPO employs a shrinkage estimator that dynamically balances local group statistics with a global prior updated via Welford's online algorithm. Theoretically, we demonstrate that EBPO guarantees strictly lower Mean Squared Error (MSE), bounded entropy decay, and non-vanishing penalty signals in failure scenarios compared to GRPO. Empirically, EBPO consistently outperforms GRPO and other established baselines across diverse benchmarks, including AIME and OlympiadBench. Notably, EBPO exhibits superior training stability, achieving high-performance gains even with small group sizes, and benefits significantly from difficulty-stratified curriculum learning.

Sandeep Singh Sengar, Abhishek Kumar, Owen Singh

This study presents significant enhancements in human pose estimation using the MediaPipe framework. The research focuses on improving accuracy, computational efficiency, and real-time processing capabilities by comprehensively optimising the underlying algorithms. Novel modifications are introduced that substantially enhance pose estimation accuracy across challenging scenarios, such as dynamic movements and partial occlusions. The improved framework is benchmarked against traditional models, demonstrating considerable precision and computational speed gains. The advancements have wide-ranging applications in augmented reality, sports analytics, and healthcare, enabling more immersive experiences, refined performance analysis, and advanced patient monitoring. The study also explores the integration of these enhancements within mobile and embedded systems, addressing the need for computational efficiency and broader accessibility. The implications of this research set a new benchmark for real-time human pose estimation technologies and pave the way for future innovations in the field. The implementation code for the paper is available at https://github.com/avhixd/Human_pose_estimation.

Taufiq Ahmed, Abhishek Kumar, Constantino Álvarez Casado et al.

Object detection models often struggle with class imbalance, where rare categories appear significantly less frequently than common ones. Existing sampling-based rebalancing strategies, such as Repeat Factor Sampling (RFS) and Instance-Aware Repeat Factor Sampling (IRFS), mitigate this issue by adjusting sample frequencies based on image and instance counts. However, these methods are based on linear adjustments, which limit their effectiveness in long-tailed distributions. This work introduces Exponentially Weighted Instance-Aware Repeat Factor Sampling (E-IRFS), an extension of IRFS that applies exponential scaling to better differentiate between rare and frequent classes. E-IRFS adjusts sampling probabilities using an exponential function applied to the geometric mean of image and instance frequencies, ensuring a more adaptive rebalancing strategy. We evaluate E-IRFS on a dataset derived from the Fireman-UAV-RGBT Dataset and four additional public datasets, using YOLOv11 object detection models to identify fire, smoke, people and lakes in emergency scenarios. The results show that E-IRFS improves detection performance by 22\% over the baseline and outperforms RFS and IRFS, particularly for rare categories. The analysis also highlights that E-IRFS has a stronger effect on lightweight models with limited capacity, as these models rely more on data sampling strategies to address class imbalance. The findings demonstrate that E-IRFS improves rare object detection in resource-constrained environments, making it a suitable solution for real-time applications such as UAV-based emergency monitoring. The code is available at: https://github.com/futurians/E-IRFS.

Kushagra Pandey, Avideep Mukherjee, Piyush Rai et al.

Diffusion probabilistic models have been shown to generate state-of-the-art results on several competitive image synthesis benchmarks but lack a low-dimensional, interpretable latent space, and are slow at generation. On the other hand, standard Variational Autoencoders (VAEs) typically have access to a low-dimensional latent space but exhibit poor sample quality. We present DiffuseVAE, a novel generative framework that integrates VAE within a diffusion model framework, and leverage this to design novel conditional parameterizations for diffusion models. We show that the resulting model equips diffusion models with a low-dimensional VAE inferred latent code which can be used for downstream tasks like controllable synthesis. The proposed method also improves upon the speed vs quality tradeoff exhibited in standard unconditional DDPM/DDIM models (for instance, FID of 16.47 vs 34.36 using a standard DDIM on the CelebA-HQ-128 benchmark using T=10 reverse process steps) without having explicitly trained for such an objective. Furthermore, the proposed model exhibits synthesis quality comparable to state-of-the-art models on standard image synthesis benchmarks like CIFAR-10 and CelebA-64 while outperforming most existing VAE-based methods. Lastly, we show that the proposed method exhibits inherent generalization to different types of noise in the conditioning signal. For reproducibility, our source code is publicly available at https://github.com/kpandey008/DiffuseVAE.

Abhishek Kumar, Harikrishna Narasimhan, Andrew Cotter

We consider a popular family of constrained optimization problems arising in machine learning that involve optimizing a non-decomposable evaluation metric with a certain thresholded form, while constraining another metric of interest. Examples of such problems include optimizing the false negative rate at a fixed false positive rate, optimizing precision at a fixed recall, optimizing the area under the precision-recall or ROC curves, etc. Our key idea is to formulate a rate-constrained optimization that expresses the threshold parameter as a function of the model parameters via the Implicit Function theorem. We show how the resulting optimization problem can be solved using standard gradient based methods. Experiments on benchmark datasets demonstrate the effectiveness of our proposed method over existing state-of-the art approaches for these problems. The code for the proposed method is available at https://github.com/google-research/google-research/tree/master/implicit_constrained_optimization .

Abhishek Kumar, Deep Ghadiyali, Sridhar Chimalakonda

Stack Overflow (SO) platform has a huge dataset of questions and answers driven by interactions between users. But the count of unanswered questions is continuously rising. This issue is common across various community Question & Answering platforms (Q&A) such as Yahoo, Quora and so on. Clustering is one of the approaches used by these communities to address this challenge. Specifically, Intent-based clustering could be leveraged to answer unanswered questions using other answered questions in the same cluster and can also improve the response time for new questions. It is here, we propose SOCluster, an approach and a tool to cluster SO questions based on intent using a graph-based clustering approach. We selected four datasets of 10k, 20k, 30k & 40k SO questions without code-snippets or images involved, and performed intent-based clustering on them. We have done a preliminary evaluation of our tool by analyzing the resultant clusters using the commonly used metrics of Silhouette coefficient, Calinkski-Harabasz Index, & Davies-Bouldin Index. We performed clustering for 8 different threshold similarity values and analyzed the intriguing trends reflected by the output clusters through the three evaluation metrics. At 90% threshold similarity, it shows the best value for the three evaluation metrics on all four datasets. The source code and tool are available for download on Github at: https://github.com/Liveitabhi/SOCluster, and the demo can be found here: https://youtu.be/uyn8ie4h3NY.

Soumya Sanyal, Janakiraman Balachandran, Naganand Yadati et al.

Developing accurate, transferable and computationally inexpensive machine learning models can rapidly accelerate the discovery and development of new materials. Some of the major challenges involved in developing such models are, (i) limited availability of materials data as compared to other fields, (ii) lack of universal descriptor of materials to predict its various properties. The limited availability of materials data can be addressed through transfer learning, while the generic representation was recently addressed by Xie and Grossman [1], where they developed a crystal graph convolutional neural network (CGCNN) that provides a unified representation of crystals. In this work, we develop a new model (MT-CGCNN) by integrating CGCNN with transfer learning based on multi-task (MT) learning. We demonstrate the effectiveness of MT-CGCNN by simultaneous prediction of various material properties such as Formation Energy, Band Gap and Fermi Energy for a wide range of inorganic crystals (46774 materials). MT-CGCNN is able to reduce the test error when employed on correlated properties by upto 8%. The model prediction has lower test error compared to CGCNN, even when the training data is reduced by 10%. We also demonstrate our model's better performance through prediction of end user scenario related to metal/non-metal classification. These results encourage further development of machine learning approaches which leverage multi-task learning to address the aforementioned challenges in the discovery of new materials. We make MT-CGCNN's source code available to encourage reproducible research.

Shuangfei Zhai, Hui Wu, Abhishek Kumar et al.

Feature pooling layers (e.g., max pooling) in convolutional neural networks (CNNs) serve the dual purpose of providing increasingly abstract representations as well as yielding computational savings in subsequent convolutional layers. We view the pooling operation in CNNs as a two-step procedure: first, a pooling window (e.g., $2\times 2$) slides over the feature map with stride one which leaves the spatial resolution intact, and second, downsampling is performed by selecting one pixel from each non-overlapping pooling window in an often uniform and deterministic (e.g., top-left) manner. Our starting point in this work is the observation that this regularly spaced downsampling arising from non-overlapping windows, although intuitive from a signal processing perspective (which has the goal of signal reconstruction), is not necessarily optimal for \emph{learning} (where the goal is to generalize). We study this aspect and propose a novel pooling strategy with stochastic spatial sampling (S3Pool), where the regular downsampling is replaced by a more general stochastic version. We observe that this general stochasticity acts as a strong regularizer, and can also be seen as doing implicit data augmentation by introducing distortions in the feature maps. We further introduce a mechanism to control the amount of distortion to suit different datasets and architectures. To demonstrate the effectiveness of the proposed approach, we perform extensive experiments on several popular image classification benchmarks, observing excellent improvements over baseline models. Experimental code is available at https://github.com/Shuangfei/s3pool.

Abhishek Kumar, Riya Tapwal, Carsten Maple et al.

Digital identity verification systems used in remote onboarding rely on document images to authenticate users, making them vulnerable to localized manipulations of key identity fields such as facial photographs and textual information. Existing forgery detection methods, developed primarily for natural-image forensics, show limited transferability to structured identity documents. We propose FLiD, a lightweight field-localized framework that targets critical identity regions rather than processing full-document images. A fine-tuned object detector first localizes face and text fields; a frozen MobileNetV3-Small backbone then extracts compact field-level embeddings, which are classified by lightweight neural network with only 191K trainable parameters. FLiD achieves AUC scores of 0.880 (face), 0.954 (text), and 0.923 (both-field attacks), with corresponding EERs of 18.05%, 11.61%, and 15.16%, representing absolute reductions of 29-35 percentage points over a full-document baseline trained from scratch. FLiD also consistently outperforms general-purpose manipulation detectors (TruFor, MMFusion, UniVAD) across all attack scenarios while requiring 13x fewer parameters and 21x fewer FLOPs

Abhishek Kumar, Riya Tapwal, Carsten Maple

Large Language Models (LLMs) are increasingly integrated into vehicle-based digital assistants, where unsafe, ambiguous, or legally incorrect responses can lead to serious safety, ethical, and regulatory consequences. Despite growing interest in LLM safety, existing taxonomies and evaluation frameworks remain largely general-purpose and fail to capture the domain-specific risks inherent to real-world driving scenarios. In this paper, we introduce DriveSafe, a hierarchical, four-level risk taxonomy designed to systematically characterize safety-critical failure modes of LLM-based driving assistants. The taxonomy comprises 129 fine-grained atomic risk categories spanning technical, legal, societal, and ethical dimensions, grounded in real-world driving regulations and safety principles and reviewed by domain experts. To validate the safety relevance and realism of the constructed prompts, we evaluate their refusal behavior across six widely deployed LLMs. Our analysis shows that the evaluated models often fail to appropriately refuse unsafe or non-compliant driving-related queries, underscoring the limitations of general-purpose safety alignment in driving contexts.

Isha Motiyani, Abhishek Kumar, Tilak Kasturi

Despite the growing prevalence of large language models (LLMs) in domain-specific applications, the challenge of query understanding in the automotive sector still remains underexplored. This domain presents unique complexities due to its specialized vocabulary and the diverse range of user intents it encompasses. Unlike general-purpose assistants, automotive systems must precisely interpret user queries and route them to appropriate underlying tool, each designed to fulfill a distinct task such as part recommendations, repair procedures, or regulatory lookups. Moreover, these systems must extract structured inputs precisely aligned with the schema required by each tool. In this study, we present a novel two-step system for domain-specific query interpretation in the automotive context that achieves an effective balance between responsiveness, reliability, and scalability. Our initial single-step approach, which jointly performed classification and entity extraction, exhibited moderate performance and higher latency. By decomposing the task into a lightweight classification stage followed by targeted entity extraction using smaller, specialized prompts, our system achieves substantial gains in both efficiency and accuracy. Due to the niche nature of the automotive domain, we also curated a high-quality dataset by combining manually annotated and synthetically generated samples, all reviewed by domain experts. Overall, our findings demonstrate that decomposing query understanding into modular subtasks leads to a scalable, accurate, and latency-efficient solution. This approach establishes a strong ground for practical deployment in real-world automotive query understanding systems.

Abhishek Kumar, Aashraya Sachdeva

Entity recognition in Automatic Speech Recognition (ASR) is challenging for rare and domain-specific terms. In domains such as finance, medicine, and air traffic control, these errors are costly. If the entities are entirely absent from the ASR output, post-ASR correction becomes difficult. To address this, we introduce RECOVER, an agentic correction framework that serves as a tool-using agent. It leverages multiple hypotheses as evidence from ASR, retrieves relevant entities, and applies Large Language Model (LLM) correction under constraints. The hypotheses are used using different strategies, namely, 1-Best, Entity-Aware Select, Recognizer Output Voting Error Reduction (ROVER) Ensemble, and LLM-Select. Evaluated across five diverse datasets, it achieves 8-46% relative reductions in entity-phrase word error rate (E-WER) and increases recall by up to 22 percentage points. The LLM-Select achieves the best overall performance in entity correction while maintaining overall WER.

Talha Zeeshan, Abhishek Kumar, Susanna Pirttikangas et al.

This paper presents the development and evaluation of a Large Language Model (LLM), also known as foundation models, based multi-agent system framework for complex event processing (CEP) with a focus on video query processing use cases. The primary goal is to create a proof-of-concept (POC) that integrates state-of-the-art LLM orchestration frameworks with publish/subscribe (pub/sub) tools to address the integration of LLMs with current CEP systems. Utilizing the Autogen framework in conjunction with Kafka message brokers, the system demonstrates an autonomous CEP pipeline capable of handling complex workflows. Extensive experiments evaluate the system's performance across varying configurations, complexities, and video resolutions, revealing the trade-offs between functionality and latency. The results show that while higher agent count and video complexities increase latency, the system maintains high consistency in narrative coherence. This research builds upon and contributes to, existing novel approaches to distributed AI systems, offering detailed insights into integrating such systems into existing infrastructures.

Yifan Yang, Kai Zhen, Bhavana Ganesh et al.

Large Language Models (LLMs) pruning seeks to remove unimportant weights for inference speedup with minimal accuracy impact. However, existing methods often suffer from accuracy degradation without full-model sparsity-aware fine-tuning. This paper presents Wanda++, a novel pruning framework that outperforms the state-of-the-art methods by utilizing decoder-block-level \textbf{regional} gradients. Specifically, Wanda++ improves the pruning score with regional gradients for the first time and proposes an efficient regional optimization method to minimize pruning-induced output discrepancies between the dense and sparse decoder output. Notably, Wanda++ improves perplexity by up to 32\% over Wanda in the language modeling task and generalizes effectively to downstream tasks. Moreover, despite updating weights with regional optimization, Wanda++ remains orthogonal to sparsity-aware fine-tuning, further reducing perplexity with LoRA in great extend. Our approach is lightweight, pruning a 7B LLaMA model in under 10 minutes on a single H100 GPU.

Burak Varıcı, Emre Acartürk, Karthikeyan Shanmugam et al.

This paper addresses intervention-based causal representation learning (CRL) under a general nonparametric latent causal model and an unknown transformation that maps the latent variables to the observed variables. Linear and general transformations are investigated. The paper addresses both the identifiability and achievability aspects. Identifiability refers to determining algorithm-agnostic conditions that ensure the recovery of the true latent causal variables and the underlying latent causal graph. Achievability refers to the algorithmic aspects and addresses designing algorithms that achieve identifiability guarantees. By drawing novel connections between score functions (i.e., the gradients of the logarithm of density functions) and CRL, this paper designs a score-based class of algorithms that ensures both identifiability and achievability. First, the paper focuses on linear transformations and shows that one stochastic hard intervention per node suffices to guarantee identifiability. It also provides partial identifiability guarantees for soft interventions, including identifiability up to mixing with parents for general causal models and perfect recovery of the latent graph for sufficiently nonlinear causal models. Secondly, it focuses on general transformations and demonstrates that two stochastic hard interventions per node are sufficient for identifiability. This is achieved by defining a differentiable loss function whose global optima ensure identifiability for general CRL. Notably, one does not need to know which pair of interventional environments has the same node intervened. Finally, the theoretical results are empirically validated via experiments on structured synthetic data and image data.

Abhishek Kumar, Sarfaroz Yunusov, Ali Emami

Research on Large Language Models (LLMs) has often neglected subtle biases that, although less apparent, can significantly influence the models' outputs toward particular social narratives. This study addresses two such biases within LLMs: representative bias, which denotes a tendency of LLMs to generate outputs that mirror the experiences of certain identity groups, and affinity bias, reflecting the models' evaluative preferences for specific narratives or viewpoints. We introduce two novel metrics to measure these biases: the Representative Bias Score (RBS) and the Affinity Bias Score (ABS), and present the Creativity-Oriented Generation Suite (CoGS), a collection of open-ended tasks such as short story writing and poetry composition, designed with customized rubrics to detect these subtle biases. Our analysis uncovers marked representative biases in prominent LLMs, with a preference for identities associated with being white, straight, and men. Furthermore, our investigation of affinity bias reveals distinctive evaluative patterns within each model, akin to `bias fingerprints'. This trend is also seen in human evaluators, highlighting a complex interplay between human and machine bias perceptions.

Mintong Kang, Vinayshekhar Bannihatti Kumar, Shamik Roy et al.

Text-to-image diffusion models often exhibit biases toward specific demographic groups, such as generating more males than females when prompted to generate images of engineers, raising ethical concerns and limiting their adoption. In this paper, we tackle the challenge of mitigating generation bias towards any target attribute value (e.g., "male" for "gender") in diffusion models while preserving generation quality. We propose FairGen, an adaptive latent guidance mechanism which controls the generation distribution during inference. In FairGen, a latent guidance module dynamically adjusts the diffusion process to enforce specific attributes, while a memory module tracks the generation statistics and steers latent guidance to align with the targeted fair distribution of the attribute values. Furthermore, we address the limitations of existing datasets by introducing the Holistic Bias Evaluation (HBE) benchmark, which covers diverse domains and incorporates complex prompts to assess bias more comprehensively. Extensive evaluations on HBE and Stable Bias datasets demonstrate that FairGen outperforms existing bias mitigation approaches, achieving substantial bias reduction (e.g., 68.5% gender bias reduction on Stable Diffusion 2). Ablation studies highlight FairGen's ability to flexibly control the output distribution at any user-specified granularity, ensuring adaptive and targeted bias mitigation.

Yuhang Zhou, Mingrui Zhang, Ke Li et al.

Understanding and reasoning over tables is a critical capability for many real-world applications. Large language models (LLMs) have shown promise on this task, but current approaches remain limited. Fine-tuning based methods strengthen language reasoning; yet they are prone to arithmetic errors and hallucination. In contrast, tool-based methods enable precise table manipulation but rely on rigid schemas and lack semantic understanding. These complementary drawbacks highlight the need for approaches that integrate robust reasoning with reliable table processing. In this work, we propose Mixture-of-Minds, a multi-agent framework that decomposes table reasoning into three specialized roles: planning, coding, and answering. This design enables each agent to focus on a specific aspect of the task while leveraging code execution for precise table manipulation. Building on this workflow, we introduce a self-improvement training framework that employs Monte Carlo Tree Search (MCTS) rollouts to generate pseudo-gold trajectories and optimize agents with reinforcement learning (RL). Extensive experiments show that Mixture-of-Minds delivers substantial gains, reaching 62.13% on TableBench and surpassing OpenAI-o4-mini-high. These results demonstrate the promise of combining structured multi-agent workflows with RL to advance table understanding.

Garima Chhikara, Abhishek Kumar, Abhijnan Chakraborty

Large Language Models (LLMs) have shown remarkable advancements but also raise concerns about cultural bias, often reflecting dominant narratives at the expense of under-represented subcultures. In this study, we evaluate the capacity of LLMs to recognize and accurately respond to the Little Traditions within Indian society, encompassing localized cultural practices and subcultures such as caste, kinship, marriage, and religion. Through a series of case studies, we assess whether LLMs can balance the interplay between dominant Great Traditions and localized Little Traditions. We explore various prompting strategies and further investigate whether using prompts in regional languages enhances the models cultural sensitivity and response quality. Our findings reveal that while LLMs demonstrate an ability to articulate cultural nuances, they often struggle to apply this understanding in practical, context-specific scenarios. To the best of our knowledge, this is the first study to analyze LLMs engagement with Indian subcultures, offering critical insights into the challenges of embedding cultural diversity in AI systems.

Cheng-Yu Hsieh, Yung-Sung Chuang, Chun-Liang Li et al.

Large language models (LLMs), even when specifically trained to process long input contexts, struggle to capture relevant information located in the middle of their input. This phenomenon has been known as the lost-in-the-middle problem. In this work, we make three contributions. First, we set out to understand the factors that cause this phenomenon. In doing so, we establish a connection between lost-in-the-middle to LLMs' intrinsic attention bias: LLMs exhibit a U-shaped attention bias where the tokens at the beginning and at the end of its input receive higher attention, regardless of their relevance. Second, we mitigate this positional bias through a calibration mechanism, found-in-the-middle, that allows the model to attend to contexts faithfully according to their relevance, even though when they are in the middle. Third, we show found-in-the-middle not only achieves better performance in locating relevant information within a long context, but also eventually leads to improved retrieval-augmented generation (RAG) performance across various tasks, outperforming existing methods by up to 15 percentage points. These findings open up future directions in understanding LLM attention bias and its potential consequences.

Giannis Daras, Wen-Sheng Chu, Abhishek Kumar et al.

We introduce a novel framework for solving inverse problems using NeRF-style generative models. We are interested in the problem of 3-D scene reconstruction given a single 2-D image and known camera parameters. We show that naively optimizing the latent space leads to artifacts and poor novel view rendering. We attribute this problem to volume obstructions that are clear in the 3-D geometry and become visible in the renderings of novel views. We propose a novel radiance field regularization method to obtain better 3-D surfaces and improved novel views given single view observations. Our method naturally extends to general inverse problems including inpainting where one observes only partially a single view. We experimentally evaluate our method, achieving visual improvements and performance boosts over the baselines in a wide range of tasks. Our method achieves $30-40\%$ MSE reduction and $15-25\%$ reduction in LPIPS loss compared to the previous state of the art.

Lik-Hang Lee, Zijun Lin, Rui Hu et al.

The metaverse, enormous virtual-physical cyberspace, has brought unprecedented opportunities for artists to blend every corner of our physical surroundings with digital creativity. This article conducts a comprehensive survey on computational arts, in which seven critical topics are relevant to the metaverse, describing novel artworks in blended virtual-physical realities. The topics first cover the building elements for the metaverse, e.g., virtual scenes and characters, auditory, textual elements. Next, several remarkable types of novel creations in the expanded horizons of metaverse cyberspace have been reflected, such as immersive arts, robotic arts, and other user-centric approaches fuelling contemporary creative outputs. Finally, we propose several research agendas: democratising computational arts, digital privacy, and safety for metaverse artists, ownership recognition for digital artworks, technological challenges, and so on. The survey also serves as introductory material for artists and metaverse technologists to begin creations in the realm of surrealistic cyberspace.

Abhishek Kumar, Ehsan Amid

Deep neural networks have shown impressive performance in supervised learning, enabled by their ability to fit well to the provided training data. However, their performance is largely dependent on the quality of the training data and often degrades in the presence of noise. We propose a principled approach for tackling label noise with the aim of assigning importance weights to individual instances and class labels. Our method works by formulating a class of constrained optimization problems that yield simple closed form updates for these importance weights. The proposed optimization problems are solved per mini-batch which obviates the need of storing and updating the weights over the full dataset. Our optimization framework also provides a theoretical perspective on existing label smoothing heuristics for addressing label noise (such as label bootstrapping). We evaluate our method on several benchmark datasets and observe considerable performance gains in the presence of label noise.