Sayan Ghosh

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao et al. · allen-ai, amazon-science

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting.

Adam Thelen, Xiaoge Zhang, Olga Fink et al.

As an emerging technology in the era of Industry 4.0, digital twin is gaining unprecedented attention because of its promise to further optimize process design, quality control, health monitoring, decision and policy making, and more, by comprehensively modeling the physical world as a group of interconnected digital models. In a two-part series of papers, we examine the fundamental role of different modeling techniques, twinning enabling technologies, and uncertainty quantification and optimization methods commonly used in digital twins. This second paper presents a literature review of key enabling technologies of digital twins, with an emphasis on uncertainty quantification, optimization methods, open source datasets and tools, major findings, challenges, and future directions. Discussions focus on current methods of uncertainty quantification and optimization and how they are applied in different dimensions of a digital twin. Additionally, this paper presents a case study where a battery digital twin is constructed and tested to illustrate some of the modeling and twinning methods reviewed in this two-part review. Code and preprocessed data for generating all the results and figures presented in the case study are available on GitHub.

Adam Thelen, Xiaoge Zhang, Olga Fink et al.

As an emerging technology in the era of Industry 4.0, digital twin is gaining unprecedented attention because of its promise to further optimize process design, quality control, health monitoring, decision and policy making, and more, by comprehensively modeling the physical world as a group of interconnected digital models. In a two-part series of papers, we examine the fundamental role of different modeling techniques, twinning enabling technologies, and uncertainty quantification and optimization methods commonly used in digital twins. This first paper presents a thorough literature review of digital twin trends across many disciplines currently pursuing this area of research. Then, digital twin modeling and twinning enabling technologies are further analyzed by classifying them into two main categories: physical-to-virtual, and virtual-to-physical, based on the direction in which data flows. Finally, this paper provides perspectives on the trajectory of digital twin technology over the next decade, and introduces a few emerging areas of research which will likely be of great use in future digital twin research. In part two of this review, the role of uncertainty quantification and optimization are discussed, a battery digital twin is demonstrated, and more perspectives on the future of digital twin are shared.

Karthik Prasad, Sayan Ghosh, Graham Cormode et al. · oxford

Cross-device Federated Learning is an increasingly popular machine learning setting to train a model by leveraging a large population of client devices with high privacy and security guarantees. However, communication efficiency remains a major bottleneck when scaling federated learning to production environments, particularly due to bandwidth constraints during uplink communication. In this paper, we formalize and address the problem of compressing client-to-server model updates under the Secure Aggregation primitive, a core component of Federated Learning pipelines that allows the server to aggregate the client updates without accessing them individually. In particular, we adapt standard scalar quantization and pruning methods to Secure Aggregation and propose Secure Indexing, a variant of Secure Aggregation that supports quantization for extreme compression. We establish state-of-the-art results on LEAF benchmarks in a secure Federated Learning setup with up to 40$\times$ compression in uplink communication with no meaningful loss in utility compared to uncompressed baselines.

Sayan Ghosh, Karthik Prasad, Xiaoliang Dai et al. · oxford

Neural network pruning is frequently used to compress over-parameterized networks by large amounts, while incurring only marginal drops in generalization performance. However, the impact of pruning on networks that have been highly optimized for efficient inference has not received the same level of attention. In this paper, we analyze the effect of pruning for computer vision, and study state-of-the-art ConvNets, such as the FBNetV3 family of models. We show that model pruning approaches can be used to further optimize networks trained through NAS (Neural Architecture Search). The resulting family of pruned models can consistently obtain better performance than existing FBNetV3 models at the same level of computation, and thus provide state-of-the-art results when trading off between computational complexity and generalization performance on the ImageNet benchmark. In addition to better generalization performance, we also demonstrate that when limited computation resources are available, pruning FBNetV3 models incur only a fraction of GPU-hours involved in running a full-scale NAS.

Shubham Dutta, Sayan Ghosh, Satyaki Bhattacharya et al.

An essential metric for the quality of a particle-identification experiment is its statistical power to discriminate between signal and background. Pulse shape discrimination (PSD) is a basic method for this purpose in many nuclear, high-energy and rare-event search experiments where scintillation detectors are used. Conventional techniques exploit the difference between decay-times of the pulses from signal and background events or pulse signals caused by different types of radiation quanta to achieve good discrimination. However, such techniques are efficient only when the total light-emission is sufficient to get a proper pulse profile. This is only possible when adequate amount of energy is deposited from recoil of the electrons or the nuclei of the scintillator materials caused by the incident particle on the detector. But, rare-event search experiments like direct search for dark matter do not always satisfy these conditions. Hence, it becomes imperative to have a method that can deliver a very efficient discrimination in these scenarios. Neural network based machine-learning algorithms have been used for classification problems in many areas of physics especially in high-energy experiments and have given better results compared to conventional techniques. We present the results of our investigations of two network based methods \viz Dense Neural Network and Recurrent Neural Network, for pulse shape discrimination and compare the same with conventional methods.

Aishwarya Sarkar, Sayan Ghosh, Nathan Tallent et al.

Large-scale Graph Neural Networks (GNNs) are typically trained by sampling a vertex's neighbors to a fixed distance. Because large input graphs are distributed, training requires frequent irregular communication that stalls forward progress. Moreover, fetched data changes with graph, graph distribution, sample and batch parameters, and caching polices. Consequently, any static prefetching method will miss crucial opportunities to adapt to different dynamic conditions. In this paper, we introduce Rudder, a software module embedded in the state-of-the-art AWS DistDGL framework, to autonomously prefetch remote nodes and minimize communication. Rudder's adaptation contrasts with both standard heuristics and traditional ML classifiers. We observe that the generative AI found in contemporary Large Language Models (LLMs) exhibits emergent properties like In-Context Learning (ICL) for zero-shot tasks, with logical multi-step reasoning. We find this behavior well-suited for adaptive control even with substantial undertraining. Evaluations using standard datasets and unseen configurations on the NERSC Perlmutter supercomputer show up to 91% improvement in end-to-end training performance over baseline DistDGL (no prefetching), and an 82% improvement over static prefetching, reducing communication by over 50%. Our code is available at https://github.com/aishwaryyasarkar/rudder-llm-agent.

Ashkan Yousefpour, Shen Guo, Ashish Shenoy et al.

The rapid progress of AI is fueled by increasingly large and computationally intensive machine learning models and datasets. As a consequence, the amount of compute used in training state-of-the-art models is exponentially increasing (doubling every 10 months between 2015 and 2022), resulting in a large carbon footprint. Federated Learning (FL) - a collaborative machine learning technique for training a centralized model using data of decentralized entities - can also be resource-intensive and have a significant carbon footprint, particularly when deployed at scale. Unlike centralized AI that can reliably tap into renewables at strategically placed data centers, cross-device FL may leverage as many as hundreds of millions of globally distributed end-user devices with diverse energy sources. Green AI is a novel and important research area where carbon footprint is regarded as an evaluation criterion for AI, alongside accuracy, convergence speed, and other metrics. In this paper, we propose the concept of Green FL, which involves optimizing FL parameters and making design choices to minimize carbon emissions consistent with competitive performance and training time. The contributions of this work are two-fold. First, we adopt a data-driven approach to quantify the carbon emissions of FL by directly measuring real-world at-scale FL tasks running on millions of phones. Second, we present challenges, guidelines, and lessons learned from studying the trade-off between energy efficiency, performance, and time-to-train in a production FL system. Our findings offer valuable insights into how FL can reduce its carbon footprint, and they provide a foundation for future research in the area of Green AI.

Yuwei Bao, Sayan Ghosh, Joyce Chai

Human communication is a collaborative process. Speakers, on top of conveying their own intent, adjust the content and language expressions by taking the listeners into account, including their knowledge background, personalities, and physical capabilities. Towards building AI agents with similar abilities in language communication, we propose Pragmatic Rational Speaker (PRS), a framework extending Rational Speech Act (RSA). The PRS attempts to learn the speaker-listener disparity and adjust the speech accordingly, by adding a light-weighted disparity adjustment layer into working memory on top of speaker's long-term memory system. By fixing the long-term memory, the PRS only needs to update its working memory to learn and adapt to different types of listeners. To validate our framework, we create a dataset that simulates different types of speaker-listener disparities in the context of referential games. Our empirical results demonstrate that the PRS is able to shift its output towards the language that listener are able to understand, significantly improve the collaborative task outcome.

Sayan Ghosh, Tejas Srinivasan, Swabha Swayamdipta · allen-ai

Human evaluation of generated language through pairwise preference judgments is pervasive. However, under common scenarios, such as when generations from a model pair are very similar, or when stochastic decoding results in large variations in generations, it results in inconsistent preference ratings. We address these challenges by introducing a meta-evaluation measure, separability, which estimates how suitable a test instance is for pairwise preference evaluation. For a candidate test instance, separability samples multiple generations from a pair of models, and measures how distinguishable the two sets of generations are. Our experiments show that instances with high separability values yield more consistent preference ratings from both human- and auto-raters. Further, the distribution of separability allows insights into which test benchmarks are more valuable for comparing models. Finally, we incorporate separability into ELO ratings, accounting for how suitable each test instance might be for reliably ranking LLMs. Overall, separability has implications for consistent, efficient and robust preference evaluation of LLMs with both human- and auto-raters.

Rakesh R Menon, Sayan Ghosh, Shashank Srivastava

Supervised learning has traditionally focused on inductive learning by observing labeled examples of a task. In contrast, humans have the ability to learn new concepts from language. Here, we explore training zero-shot classifiers for structured data purely from language. For this, we introduce CLUES, a benchmark for Classifier Learning Using natural language ExplanationS, consisting of a range of classification tasks over structured data along with natural language supervision in the form of explanations. CLUES consists of 36 real-world and 144 synthetic classification tasks. It contains crowdsourced explanations describing real-world tasks from multiple teachers and programmatically generated explanations for the synthetic tasks. To model the influence of explanations in classifying an example, we develop ExEnt, an entailment-based model that learns classifiers using explanations. ExEnt generalizes up to 18% better (relative) on novel tasks than a baseline that does not use explanations. We delineate key challenges for automated learning from explanations, addressing which can lead to progress on CLUES in the future. Code and datasets are available at: https://clues-benchmark.github.io.

Kangda Wei, Sayan Ghosh, Rakesh R. Menon et al.

Recent approaches have explored language-guided classifiers capable of classifying examples from novel tasks when provided with task-specific natural language explanations, instructions or prompts (Sanh et al., 2022; R. Menon et al., 2022). While these classifiers can generalize in zero-shot settings, their task performance often varies substantially between different language explanations in unpredictable ways (Lu et al., 2022; Gonen et al., 2022). Also, current approaches fail to leverage unlabeled examples that may be available in many scenarios. Here, we introduce TALC, a framework that uses data programming to adapt a language-guided classifier for a new task during inference when provided with explanations from multiple teachers and unlabeled test examples. Our results show that TALC consistently outperforms a competitive baseline from prior work by an impressive 9.3% (relative improvement). Further, we demonstrate the robustness of TALC to variations in the quality and quantity of provided explanations, highlighting its potential in scenarios where learning from multiple teachers or a crowd is involved. Our code is available at: https://github.com/WeiKangda/TALC.git.

Julia Mendelsohn, Sayan Ghosh, David Jurgens et al.

Social media enables the rapid spread of many kinds of information, from memes to social movements. However, little is known about how information crosses linguistic boundaries. We apply causal inference techniques on the European Twitter network to quantify multilingual users' structural role and communication influence in cross-lingual information exchange. Overall, multilinguals play an essential role; posting in multiple languages increases betweenness centrality by 13%, and having a multilingual network neighbor increases monolinguals' odds of sharing domains and hashtags from another language 16-fold and 4-fold, respectively. We further show that multilinguals have a greater impact on diffusing information less accessible to their monolingual compatriots, such as information from far-away countries and content about regional politics, nascent social movements, and job opportunities. By highlighting information exchange across borders, this work sheds light on a crucial component of how information and ideas spread around the world.

Cunyang Wei, Siddharth Singh, Aishwarya Sarkar et al.

Graph neural networks (GNNs) are widely used for learning on graph datasets derived from various real-world scenarios. Learning from extremely large graphs requires distributed training, and mini-batching with sampling is a popular approach for parallelizing GNN training. Existing distributed mini-batch approaches have significant performance bottlenecks due to expensive sampling methods and limited scaling when using data parallelism. In this work, we present ScaleGNN, a 4D parallel framework for scalable mini-batch GNN training that combines communication-free distributed sampling, 3D parallel matrix multiplication (PMM), and data parallelism. ScaleGNN introduces a uniform vertex sampling algorithm, enabling each process (GPU device) to construct its local mini-batch, i.e., subgraph partitions without any inter-process communication. 3D PMM enables scaling mini-batch training to much larger GPU counts than vanilla data parallelism with significantly lower communication overheads. We also present additional optimizations to overlap sampling with training, reduce communication overhead by sending data in lower precision, kernel fusion, and communication-computation overlap. We evaluate ScaleGNN on five graph datasets and demonstrate strong scaling up to 2048 GPUs on Perlmutter, 2048 GCDs on Frontier, and 1024 GPUs on Tuolumne. On Perlmutter, ScaleGNN achieves 3.5x end-to-end training speedup over the SOTA baseline on ogbn-products.

Sayan Ghosh, Rakesh R Menon, Shashank Srivastava

A hallmark of human intelligence is the ability to learn new concepts purely from language. Several recent approaches have explored training machine learning models via natural language supervision. However, these approaches fall short in leveraging linguistic quantifiers (such as 'always' or 'rarely') and mimicking humans in compositionally learning complex tasks. Here, we present LaSQuE, a method that can learn zero-shot classifiers from language explanations by using three new strategies - (1) modeling the semantics of linguistic quantifiers in explanations (including exploiting ordinal strength relationships, such as 'always' > 'likely'), (2) aggregating information from multiple explanations using an attention-based mechanism, and (3) model training via curriculum learning. With these strategies, LaSQuE outperforms prior work, showing an absolute gain of up to 7% in generalizing to unseen real-world classification tasks.

Jessica Zhao, Sayan Ghosh, Akash Bharadwaj et al.

Semi-Supervised Learning (SSL) has received extensive attention in the domain of computer vision, leading to development of promising approaches such as FixMatch. In scenarios where training data is decentralized and resides on client devices, SSL must be integrated with privacy-aware training techniques such as Federated Learning. We consider the problem of federated image classification and study the performance and privacy challenges with existing federated SSL (FSSL) approaches. Firstly, we note that even state-of-the-art FSSL algorithms can trivially compromise client privacy and other real-world constraints such as client statelessness and communication cost. Secondly, we observe that it is challenging to integrate EMA (Exponential Moving Average) updates into the federated setting, which comes at a trade-off between performance and communication cost. We propose a novel approach FedSwitch, that improves privacy as well as generalization performance through Exponential Moving Average (EMA) updates. FedSwitch utilizes a federated semi-supervised teacher-student EMA framework with two features - local teacher adaptation and adaptive switching between teacher and student for pseudo-label generation. Our proposed approach outperforms the state-of-the-art on federated image classification, can be adapted to real-world constraints, and achieves good generalization performance with minimal communication cost overhead.

Yigitcan Comlek, Sandipp Krishnan Ravi, Piyush Pandita et al.

Artificial intelligence and machine learning frameworks have served as computationally efficient mapping between inputs and outputs for engineering problems. These mappings have enabled optimization and analysis routines that have warranted superior designs, ingenious material systems and optimized manufacturing processes. A common occurrence in such modeling endeavors is the existence of multiple source of data, each differentiated by fidelity, operating conditions, experimental conditions, and more. Data fusion frameworks have opened the possibility of combining such differentiated sources into single unified models, enabling improved accuracy and knowledge transfer. However, these frameworks encounter limitations when the different sources are heterogeneous in nature, i.e., not sharing the same input parameter space. These heterogeneous input scenarios can occur when the domains differentiated by complexity, scale, and fidelity require different parametrizations. Towards addressing this void, a heterogeneous multi-source data fusion framework is proposed based on input mapping calibration (IMC) and latent variable Gaussian process (LVGP). In the first stage, the IMC algorithm is utilized to transform the heterogeneous input parameter spaces into a unified reference parameter space. In the second stage, a multi-source data fusion model enabled by LVGP is leveraged to build a single source-aware surrogate model on the transformed reference space. The proposed framework is demonstrated and analyzed on three engineering case studies (design of cantilever beam, design of ellipsoidal void and modeling properties of Ti6Al4V alloy). The results indicate that the proposed framework provides improved predictive accuracy over a single source model and transformed but source unaware model.

Yating Wu, Yuhao Zhang, Sayan Ghosh et al.

Large language model (LLM) agents often struggle in long-context interactions. As the agent accumulates more interaction history, context management approaches such as sliding window and prompt compression may omit earlier structured information that later steps rely on. Recent retrieval-based memory systems surface relevant content but still overlook the causal and logical structure needed for multi-step reasoning. We introduce ContextWeaver, a selective and dependency-structured memory framework that organizes an agent's interaction trace into a graph of reasoning steps and selects the relevant context for future actions. Unlike prior context management approaches, ContextWeaver supports: (1) dependency-based construction and traversal that link each step to the earlier steps it relies on; (2) compact dependency summarization that condenses root-to-step reasoning paths into reusable units; and (3) a lightweight validation layer that incorporates execution feedback. On the SWE-Bench Verified and Lite benchmarks, ContextWeaver improves performance over a sliding-window baseline in pass@1, while reducing reasoning steps and token usage. Our observations suggest that modeling logical dependencies provides a stable and scalable memory mechanism for LLM agents that use tools.

Yiyuan Li, Rakesh R. Menon, Sayan Ghosh et al.

Generalized quantifiers (e.g., few, most) are used to indicate the proportions predicates are satisfied (for example, some apples are red). One way to interpret quantifier semantics is to explicitly bind these satisfactions with percentage scopes (e.g., 30%-40% of apples are red). This approach can be helpful for tasks like logic formalization and surface-form quantitative reasoning (Gordon and Schubert, 2010; Roy et al., 2015). However, it remains unclear if recent foundation models possess this ability, as they lack direct training signals. To explore this, we introduce QuRe, a crowd-sourced dataset of human-annotated generalized quantifiers in Wikipedia sentences featuring percentage-equipped predicates. We explore quantifier comprehension in language models using PRESQUE, a framework that combines natural language inference and the Rational Speech Acts framework. Experimental results on the HVD dataset and QuRe illustrate that PRESQUE, employing pragmatic reasoning, performs 20% better than a literal reasoning baseline when predicting quantifier percentage scopes, with no additional training required.

Ashkan Yousefpour, Igor Shilov, Alexandre Sablayrolles et al.

We introduce Opacus, a free, open-source PyTorch library for training deep learning models with differential privacy (hosted at opacus.ai). Opacus is designed for simplicity, flexibility, and speed. It provides a simple and user-friendly API, and enables machine learning practitioners to make a training pipeline private by adding as little as two lines to their code. It supports a wide variety of layers, including multi-head attention, convolution, LSTM, GRU (and generic RNN), and embedding, right out of the box and provides the means for supporting other user-defined layers. Opacus computes batched per-sample gradients, providing higher efficiency compared to the traditional "micro batch" approach. In this paper we present Opacus, detail the principles that drove its implementation and unique features, and benchmark it against other frameworks for training models with differential privacy as well as standard PyTorch.

Swarnadeep Saha, Sayan Ghosh, Shashank Srivastava et al.

Recent work by Clark et al. (2020) shows that transformers can act as 'soft theorem provers' by answering questions over explicitly provided knowledge in natural language. In our work, we take a step closer to emulating formal theorem provers, by proposing PROVER, an interpretable transformer-based model that jointly answers binary questions over rule-bases and generates the corresponding proofs. Our model learns to predict nodes and edges corresponding to proof graphs in an efficient constrained training paradigm. During inference, a valid proof, satisfying a set of global constraints is generated. We conduct experiments on synthetic, hand-authored, and human-paraphrased rule-bases to show promising results for QA and proof generation, with strong generalization performance. First, PROVER generates proofs with an accuracy of 87%, while retaining or improving performance on the QA task, compared to RuleTakers (up to 6% improvement on zero-shot evaluation). Second, when trained on questions requiring lower depths of reasoning, it generalizes significantly better to higher depths (up to 15% improvement). Third, PROVER obtains near perfect QA accuracy of 98% using only 40% of the training data. However, generating proofs for questions requiring higher depths of reasoning becomes challenging, and the accuracy drops to 65% for 'depth 5', indicating significant scope for future work. Our code and models are publicly available at https://github.com/swarnaHub/PRover

Abrar Anwar, Rohan Gupta, Zain Merchant et al.

Evaluating learned robot control policies to determine their physical task-level capabilities costs experimenter time and effort. The growing number of policies and tasks exacerbates this issue. It is impractical to test every policy on every task multiple times; each trial requires a manual environment reset, and each task change involves re-arranging objects or even changing robots. Naively selecting a random subset of tasks and policies to evaluate is a high-cost solution with unreliable, incomplete results. In this work, we formulate robot evaluation as an active testing problem. We propose to model the distribution of robot performance across all tasks and policies as we sequentially execute experiments. Tasks often share similarities that can reveal potential relationships in policy behavior, and we show that natural language is a useful prior in modeling these relationships between tasks. We then leverage this formulation to reduce the experimenter effort by using a cost-aware expected information gain heuristic to efficiently select informative trials. Our framework accommodates both continuous and discrete performance outcomes. We conduct experiments on existing evaluation data from real robots and simulations. By prioritizing informative trials, our framework reduces the cost of calculating evaluation metrics for robot policies across many tasks.

Sandipp Krishnan Ravi, Yigitcan Comlek, Arjun Pathak et al.

With the advent of artificial intelligence and machine learning, various domains of science and engineering communities have leveraged data-driven surrogates to model complex systems through fusing numerous sources of information (data) from published papers, patents, open repositories, or other resources. However, not much attention has been paid to the differences in quality and comprehensiveness of the known and unknown underlying physical parameters of the information sources, which could have downstream implications during system optimization. Additionally, existing methods cannot fuse multi-source data into a single predictive model. Towards resolving this issue, a multi-source data fusion framework based on Latent Variable Gaussian Process (LVGP) is proposed. The individual data sources are tagged as a characteristic categorical variable that are mapped into a physically interpretable latent space, allowing the development of source-aware data fusion modeling. Additionally, a dissimilarity metric based on the latent variables of LVGP is introduced to study and understand the differences in the sources of data. The proposed approach is demonstrated on and analyzed through two mathematical and two materials science case studies. From the case studies, it is observed that compared to using single-source and source unaware machine learning models, the proposed multi-source data fusion framework can provide better predictions for sparse-data problems.

Aishwarya Sarkar, Sayan Ghosh, Nathan R. Tallent et al.

Graph Neural Networks (GNN) are indispensable in learning from graph-structured data, yet their rising computational costs, especially on massively connected graphs, pose significant challenges in terms of execution performance. To tackle this, distributed-memory solutions such as partitioning the graph to concurrently train multiple replicas of GNNs are in practice. However, approaches requiring a partitioned graph usually suffer from communication overhead and load imbalance, even under optimal partitioning and communication strategies due to irregularities in the neighborhood minibatch sampling. This paper proposes practical trade-offs for improving the sampling and communication overheads for representation learning on distributed graphs (using popular GraphSAGE architecture) by developing a parameterized continuous prefetch and eviction scheme on top of the state-of-the-art Amazon DistDGL distributed GNN framework, demonstrating about 15-40% improvement in end-to-end training performance on the National Energy Research Scientific Computing Center's (NERSC) Perlmutter supercomputer for various OGB datasets.

Aishwarya Sarkar, Sayan Ghosh, Nathan R. Tallent et al.

Successful machine learning on graphs or networks requires embeddings that not only represent nodes and edges as low-dimensional vectors but also preserve the graph structure. Established methods for generating embeddings require flexible exploration of the entire graph through repeated use of random walks that capture graph structure with samples of nodes and edges. These methods create scalability challenges for massive graphs with millions-to-billions of edges because single-node solutions have inadequate memory and processing capabilities. We present NOMAD, a distributed-memory graph embedding framework using the Message Passing Interface (MPI) for distributed graphs. NOMAD implements proximity-based models proposed in the widely popular LINE (Large-scale Information Network Embedding) algorithm. We propose several practical trade-offs to improve the scalability and communication overheads confronted by irregular and distributed graph embedding methods, catering to massive-scale graphs arising in web and science domains. NOMAD demonstrates median speedups of 10/100x on CPU-based NERSC Perlmutter cluster relative to the popular reference implementations of multi-threaded LINE and node2vec, 35-76x over distributed PBG, and competitive embedding quality relative to LINE, node2vec, and GraphVite, while yielding 12-370x end-to-end speedups on real-world graphs.

Nathan Tallent, Jan Strube, Luanzheng Guo et al.

Automating the theory-experiment cycle requires effective distributed workflows that utilize a computing continuum spanning lab instruments, edge sensors, computing resources at multiple facilities, data sets distributed across multiple information sources, and potentially cloud. Unfortunately, the obvious methods for constructing continuum platforms, orchestrating workflow tasks, and curating datasets over time fail to achieve scientific requirements for performance, energy, security, and reliability. Furthermore, achieving the best use of continuum resources depends upon the efficient composition and execution of workflow tasks, i.e., combinations of numerical solvers, data analytics, and machine learning. Pacific Northwest National Laboratory's LDRD "Cloud, High-Performance Computing (HPC), and Edge for Science and Security" (CHESS) has developed a set of interrelated capabilities for enabling distributed scientific workflows and curating datasets. This report describes the results and successes of CHESS from the perspective of open science.

Ashish S. Nair, Sandipp Krishnan Ravi, Itzel Salgado et al.

Generative AI has emerged as a transformative paradigm in engineering design, enabling automated synthesis and reconstruction of complex 3D geometries while preserving feasibility and performance relevance. This paper introduces a domain-specific implicit generative framework for turbine blade geometry using DeepSDF, addressing critical gaps in performance-aware modeling and manufacturable design generation. The proposed method leverages a continuous signed distance function (SDF) representation to reconstruct and generate smooth, watertight geometries with quantified accuracy. It establishes an interpretable, near-Gaussian latent space that aligns with blade-relevant parameters, such as taper and chord ratios, enabling controlled exploration and unconditional synthesis through interpolation and Gaussian sampling. In addition, a compact neural network maps engineering descriptors, such as maximum directional strains, to latent codes, facilitating the generation of performance-informed geometry. The framework achieves high reconstruction fidelity, with surface distance errors concentrated within $1\%$ of the maximum blade dimension, and demonstrates robust generalization to unseen designs. By integrating constraints, objectives, and performance metrics, this approach advances beyond traditional 2D-guided or unconstrained 3D pipelines, offering a practical and interpretable solution for data-driven turbine blade modeling and concept generation.

Guoyao Li, Ran He, Shusen Jing et al.

Large language models (LLMs) excel at capturing semantic nuances and therefore show impressive relevance ranking performance in modern recommendation and search systems. However, they suffer from high computational overhead under industrial latency and throughput requirements. In particular, cross-encoder ranking systems often create long context prefill-heavy workloads, as the model has to be presented with the user, query and item information. To this end, we propose MixLM, a novel LLM-based ranking framework, which significantly improves the system throughput via reducing the input context length, while preserving the semantic strength of cross-encoder rankers. In contrast to a standard ranking system where the context is presented to the model as pure text, we propose to use mix-interaction, a mixture of text and embedding tokens to represent the input. Specifically, MixLM encodes all items in the catalog into a few embedding tokens and stores in a nearline cache. The encoded item descriptions are used during online inference, effectively reducing the item length from a few thousand text tokens to a few embedding tokens. We share insights from deploying our MixLM framework to a real-world search application at LinkedIn, including a detailed discussion of our training pipelines, as well as a thorough analysis of our online serving infrastructure optimization. With the same latency budget and on-par relevance metrics, MixLM increased throughput by 10.0x comparing with strong baselines, 75.9x over full-text LLM rerankers. The efficiency gains delivered by MixLM enabled full-traffic deployment of LLM-powered search, which resulted in a significant 0.47\% increase in Daily Active Users (DAU) in online A/B tests.

Kayhan Behdin, Qingquan Song, Sriram Vasudevan et al.

Large Language Models (LLMs) have demonstrated impressive quality when applied to predictive tasks such as relevance ranking and semantic search. However, deployment of such LLMs remains prohibitively expensive for industry applications with strict latency and throughput requirements. In this work, we present lessons and efficiency insights from developing a purely text-based decoder-only Small Language Model (SLM) for a semantic search application at LinkedIn. Particularly, we discuss model compression techniques such as pruning that allow us to reduce the model size by up to $40\%$ while maintaining the accuracy. Additionally, we present context compression techniques that allow us to reduce the input context length by up to $10$x with minimal loss of accuracy. Finally, we present practical lessons from optimizing the serving infrastructure for deploying such a system on GPUs at scale, serving millions of requests per second. Taken together, this allows us to increase our system's throughput by $10$x in a real-world deployment, while meeting our quality bar.

Sayan Ghosh, Shahzaib Saqib Warraich, Dhruv Tarsadiya et al. · allen-ai

Language models can be sampled multiple times to access the distribution underlying their responses, but existing methods cannot efficiently synthesize rich epistemic signals across different long-form responses. We introduce Consensus Graphs (ConGrs), a flexible DAG-based data structure that represents shared information, as well as semantic variation in a set of sampled LM responses to the same prompt. We construct ConGrs using a light-weight lexical sequence alignment algorithm from bioinformatics, supplemented by the targeted usage of a secondary LM judge. Further, we design task-dependent decoding methods to synthesize a single, final response from our ConGr data structure. Our experiments show that synthesizing responses from ConGrs improves factual precision on two biography generation tasks by up to 31% over an average response and reduces reliance on LM judges by more than 80% compared to other methods. We also use ConGrs for three refusal-based tasks requiring abstention on unanswerable queries and find that abstention rate is increased by up to 56%. We apply our approach to the MATH and AIME reasoning tasks and find an improvement over self-verification and majority vote baselines by up to 6 points of accuracy. We show that ConGrs provide a flexible method for capturing variation in LM responses and using the epistemic signals provided by response variation to synthesize more effective responses.

Wenhui Zhu, Xiwen Chen, Zhipeng Wang et al.

Instructed Visual Segmentation (IVS) tasks require segmenting objects in images or videos based on natural language instructions. While recent multimodal large language models (MLLMs) have achieved strong performance on IVS, their inference cost remains a major bottleneck, particularly in video. We empirically analyze visual token sampling in MLLMs and observe a strong correlation between subset token coverage and segmentation performance. This motivates our design of a simple and effective token pruning method that selects a compact yet spatially representative subset of tokens to accelerate inference. In this paper, we introduce a novel visual token pruning method for IVS, called EVTP-IV, which builds upon the k-center by integrating spatial information to ensure better coverage. We further provide an information-theoretic analysis to support our design. Experiments on standard IVS benchmarks show that our method achieves up to 5X speed-up on video tasks and 3.5X on image tasks, while maintaining comparable accuracy using only 20% of the tokens. Our method also consistently outperforms state-of-the-art pruning baselines under varying pruning ratios.

Nurani Rajagopal Rohan, Vigneswaran C, Sayan Ghosh et al.

We propose a novel, brain-inspired deep neural network model known as the Deep Oscillatory Neural Network (DONN). Deep neural networks like the Recurrent Neural Networks indeed possess sequence processing capabilities but the internal states of the network are not designed to exhibit brain-like oscillatory activity. With this motivation, the DONN is designed to have oscillatory internal dynamics. Neurons of the DONN are either nonlinear neural oscillators or traditional neurons with sigmoidal or ReLU activation. The neural oscillator used in the model is the Hopf oscillator, with the dynamics described in the complex domain. Input can be presented to the neural oscillator in three possible modes. The sigmoid and ReLU neurons also use complex-valued extensions. All the weight stages are also complex-valued. Training follows the general principle of weight change by minimizing the output error and therefore has an overall resemblance to complex backpropagation. A generalization of DONN to convolutional networks known as the Oscillatory Convolutional Neural Network is also proposed. The two proposed oscillatory networks are applied to a variety of benchmark problems in signal and image/video processing. The performance of the proposed models is either comparable or superior to published results on the same data sets.

Bingsheng Yao, Ishan Jindal, Lucian Popa et al.

Real-world domain experts (e.g., doctors) rarely annotate only a decision label in their day-to-day workflow without providing explanations. Yet, existing low-resource learning techniques, such as Active Learning (AL), that aim to support human annotators mostly focus on the label while neglecting the natural language explanation of a data point. This work proposes a novel AL architecture to support experts' real-world need for label and explanation annotations in low-resource scenarios. Our AL architecture leverages an explanation-generation model to produce explanations guided by human explanations, a prediction model that utilizes generated explanations toward prediction faithfully, and a novel data diversity-based AL sampling strategy that benefits from the explanation annotations. Automated and human evaluations demonstrate the effectiveness of incorporating explanations into AL sampling and the improved human annotation efficiency and trustworthiness with our AL architecture. Additional ablation studies illustrate the potential of our AL architecture for transfer learning, generalizability, and integration with large language models (LLMs). While LLMs exhibit exceptional explanation-generation capabilities for relatively simple tasks, their effectiveness in complex real-world tasks warrants further in-depth study.

Yonatan Ashenafi, Piyush Pandita, Sayan Ghosh

Engineering problems that are modeled using sophisticated mathematical methods or are characterized by expensive-to-conduct tests or experiments, are encumbered with limited budget or finite computational resources. Moreover, practical scenarios in the industry, impose restrictions, based on logistics and preference, on the manner in which the experiments can be conducted. For example, material supply may enable only a handful of experiments in a single-shot or in the case of computational models one may face significant wait-time based on shared computational resources. In such scenarios, one usually resorts to performing experiments in a manner that allows for maximizing one's state-of-knowledge while satisfying the above mentioned practical constraints. Sequential design of experiments (SDOE) is a popular suite of methods, that has yielded promising results in recent years across different engineering and practical problems. A common strategy, that leverages Bayesian formalism is the Bayesian SDOE, which usually works best in the one-step-ahead or myopic scenario of selecting a single experiment at each step of a sequence of experiments. In this work, we aim to extend the SDOE strategy, to query the experiment or computer code at a batch of inputs. To this end, we leverage deep reinforcement learning (RL) based policy gradient methods, to propose batches of queries that are selected taking into account entire budget in hand. The algorithm retains the sequential nature, inherent in the SDOE, while incorporating elements of reward based on task from the domain of deep RL. A unique capability of the proposed methodology is its ability to be applied to multiple tasks, for example optimization of a function, once its trained. We demonstrate the performance of the proposed algorithm on a synthetic problem, and a challenging high-dimensional engineering problem.

Sayan Ghosh, Shashank Srivastava

Mapping natural language instructions to programs that computers can process is a fundamental challenge. Existing approaches focus on likelihood-based training or using reinforcement learning to fine-tune models based on a single reward. In this paper, we pose program generation from language as Inverse Reinforcement Learning. We introduce several interpretable reward components and jointly learn (1) a reward function that linearly combines them, and (2) a policy for program generation. Fine-tuning with our approach achieves significantly better performance than competitive methods using Reinforcement Learning (RL). On the VirtualHome framework, we get improvements of up to 9.0% on the Longest Common Subsequence metric and 14.7% on recall-based metrics over previous work on this framework (Puig et al., 2018). The approach is data-efficient, showing larger gains in performance in the low-data regime. Generated programs are also preferred by human evaluators over an RL-based approach, and rated higher on relevance, completeness, and human-likeness.

Somnath Basu Roy Chowdhury, Sayan Ghosh, Yiyuan Li et al.

Contextual representations learned by language models can often encode undesirable attributes, like demographic associations of the users, while being trained for an unrelated target task. We aim to scrub such undesirable attributes and learn fair representations while maintaining performance on the target task. In this paper, we present an adversarial learning framework "Adversarial Scrubber" (ADS), to debias contextual representations. We perform theoretical analysis to show that our framework converges without leaking demographic information under certain conditions. We extend previous evaluation techniques by evaluating debiasing performance using Minimum Description Length (MDL) probing. Experimental evaluations on 8 datasets show that ADS generates representations with minimal information about demographic attributes while being maximally informative about the target task.

Sayan Ghosh, Shashank Srivastava

While large language models have shown exciting progress on several NLP benchmarks, evaluating their ability for complex analogical reasoning remains under-explored. Here, we introduce a high-quality crowdsourced dataset of narratives for employing proverbs in context as a benchmark for abstract language understanding. The dataset provides fine-grained annotation of aligned spans between proverbs and narratives, and contains minimal lexical overlaps between narratives and proverbs, ensuring that models need to go beyond surface-level reasoning to succeed. We explore three tasks: (1) proverb recommendation and alignment prediction, (2) narrative generation for a given proverb and topic, and (3) identifying narratives with similar motifs. Our experiments show that neural language models struggle on these tasks compared to humans, and these tasks pose multiple learning challenges.

Sayan Ghosh, Govinda A. Padmanabha, Cheng Peng et al.

One of the critical components in Industrial Gas Turbines (IGT) is the turbine blade. Design of turbine blades needs to consider multiple aspects like aerodynamic efficiency, durability, safety and manufacturing, which make the design process sequential and iterative.The sequential nature of these iterations forces a long design cycle time, ranging from several months to years. Due to the reactionary nature of these iterations, little effort has been made to accumulate data in a manner that allows for deep exploration and understanding of the total design space. This is exemplified in the process of designing the individual components of the IGT resulting in a potential unrealized efficiency. To overcome the aforementioned challenges, we demonstrate a probabilistic inverse design machine learning framework (PMI), to carry out an explicit inverse design. PMI calculates the design explicitly without excessive costly iteration and overcomes the challenges associated with ill-posed inverse problems. In this work, the framework will be demonstrated on inverse aerodynamic design of three-dimensional turbine blades.

Sayan Ghosh, Dylan Baker, David Jurgens et al.

Online social media platforms increasingly rely on Natural Language Processing (NLP) techniques to detect abusive content at scale in order to mitigate the harms it causes to their users. However, these techniques suffer from various sampling and association biases present in training data, often resulting in sub-par performance on content relevant to marginalized groups, potentially furthering disproportionate harms towards them. Studies on such biases so far have focused on only a handful of axes of disparities and subgroups that have annotations/lexicons available. Consequently, biases concerning non-Western contexts are largely ignored in the literature. In this paper, we introduce a weakly supervised method to robustly detect lexical biases in broader geocultural contexts. Through a case study on a publicly available toxicity detection model, we demonstrate that our method identifies salient groups of cross-geographic errors, and, in a follow up, demonstrate that these groupings reflect human judgments of offensive and inoffensive language in those geographic contexts. We also conduct analysis of a model trained on a dataset with ground truth labels to better understand these biases, and present preliminary mitigation experiments.

Waad Subber, Piyush Pandita, Sayan Ghosh et al.

Most common mechanistic models are traditionally presented in mathematical forms to explain a given physical phenomenon. Machine learning algorithms, on the other hand, provide a mechanism to map the input data to output without explicitly describing the underlying physical process that generated the data. We propose a Data-based Physics Discovery (DPD) framework for automatic discovery of governing equations from observed data. Without a prior definition of the model structure, first a free-form of the equation is discovered, and then calibrated and validated against the available data. In addition to the observed data, the DPD framework can utilize available prior physical models, and domain expert feedback. When prior models are available, the DPD framework can discover an additive or multiplicative correction term represented symbolically. The correction term can be a function of the existing input variable to the prior model, or a newly introduced variable. In case a prior model is not available, the DPD framework discovers a new data-based standalone model governing the observations. We demonstrate the performance of the proposed framework on a real-world application in the aerospace industry.

Waad Subber, Sayan Ghosh, Piyush Pandita et al.

Industrial dynamical systems often exhibit multi-scale response due to material heterogeneities, operation conditions and complex environmental loadings. In such problems, it is the case that the smallest length-scale of the systems dynamics controls the numerical resolution required to effectively resolve the embedded physics. In practice however, high numerical resolutions is only required in a confined region of the system where fast dynamics or localized material variability are exhibited, whereas a coarser discretization can be sufficient in the rest majority of the system. To this end, a unified computational scheme with uniform spatio-temporal resolutions for uncertainty quantification can be very computationally demanding. Partitioning the complex dynamical system into smaller easier-to-solve problems based of the localized dynamics and material variability can reduce the overall computational cost. However, identifying the region of interest for high-resolution and intensive uncertainty quantification can be a problem dependent. The region of interest can be specified based on the localization features of the solution, user interest, and correlation length of the random material properties. For problems where a region of interest is not evident, Bayesian inference can provide a feasible solution. In this work, we employ a Bayesian framework to update our prior knowledge on the localized region of interest using measurements and system response. To address the computational cost of the Bayesian inference, we construct a Gaussian process surrogate for the forward model. Once, the localized region of interest is identified, we use polynomial chaos expansion to propagate the localization uncertainty. We demonstrate our framework through numerical experiments on a three-dimensional elastodynamic problem.

Raphael Gautier, Piyush Pandita, Sayan Ghosh et al.

Modern day engineering problems are ubiquitously characterized by sophisticated computer codes that map parameters or inputs to an underlying physical process. In other situations, experimental setups are used to model the physical process in a laboratory, ensuring high precision while being costly in materials and logistics. In both scenarios, only limited amount of data can be generated by querying the expensive information source at a finite number of inputs or designs. This problem is compounded further in the presence of a high-dimensional input space. State-of-the-art parameter space dimension reduction methods, such as active subspace, aim to identify a subspace of the original input space that is sufficient to explain the output response. These methods are restricted by their reliance on gradient evaluations or copious data, making them inadequate to expensive problems without direct access to gradients. The proposed methodology is gradient-free and fully Bayesian, as it quantifies uncertainty in both the low-dimensional subspace and the surrogate model parameters. This enables a full quantification of epistemic uncertainty and robustness to limited data availability. It is validated on multiple datasets from engineering and science and compared to two other state-of-the-art methods based on four aspects: a) recovery of the active subspace, b) deterministic prediction accuracy, c) probabilistic prediction accuracy, and d) training time. The comparison shows that the proposed method improves the active subspace recovery and predictive accuracy, in both the deterministic and probabilistic sense, when only few model observations are available for training, at the cost of increased training time.

Panagiotis Tsilifis, Piyush Pandita, Sayan Ghosh et al.

We present a Bayesian approach to identify optimal transformations that map model input points to low dimensional latent variables. The "projection" mapping consists of an orthonormal matrix that is considered a priori unknown and needs to be inferred jointly with the GP parameters, conditioned on the available training data. The proposed Bayesian inference scheme relies on a two-step iterative algorithm that samples from the marginal posteriors of the GP parameters and the projection matrix respectively, both using Markov Chain Monte Carlo (MCMC) sampling. In order to take into account the orthogonality constraints imposed on the orthonormal projection matrix, a Geodesic Monte Carlo sampling algorithm is employed, that is suitable for exploiting probability measures on manifolds. We extend the proposed framework to multi-fidelity models using GPs including the scenarios of training multiple outputs together. We validate our framework on three synthetic problems with a known lower-dimensional subspace. The benefits of our proposed framework, are illustrated on the computationally challenging three-dimensional aerodynamic optimization of a last-stage blade for an industrial gas turbine, where we study the effect of an 85-dimensional airfoil shape parameterization on two output quantities of interest, specifically on the aerodynamic efficiency and the degree of reaction.

Rakesh Bal, Sayan Sinha, Swastika Dutta et al.

Twitter has become one of the most sought after places to discuss a wide variety of topics, including medically relevant issues such as cancer. This helps spread awareness regarding the various causes, cures and prevention methods of cancer. However, no proper analysis has been performed, which discusses the validity of such claims. In this work, we aim to tackle the misinformation spread in such platforms. We collect and present a dataset regarding tweets which talk specifically about cancer and propose an attention-based deep learning model for automated detection of misinformation along with its spread. We then do a comparative analysis of the linguistic variation in the text corresponding to misinformation and truth. This analysis helps us gather relevant insights on various social aspects related to misinformed tweets.

Sayan Ghosh, Piyush Pandita, Steven Atkinson et al.

Industrial applications frequently pose a notorious challenge for state-of-the-art methods in the contexts of optimization, designing experiments and modeling unknown physical response. This problem is aggravated by limited availability of clean data, uncertainty in available physics-based models and additional logistic and computational expense associated with experiments. In such a scenario, Bayesian methods have played an impactful role in alleviating the aforementioned obstacles by quantifying uncertainty of different types under limited resources. These methods, usually deployed as a framework, allows decision makers to make informed choices under uncertainty while being able to incorporate information on the the fly, usually in the form of data, from multiple sources while being consistent with the physical intuition about the problem. This is a major advantage that Bayesian methods bring to fruition especially in the industrial context. This paper is a compendium of the Bayesian modeling methodology that is being consistently developed at GE Research. The methodology, called GE's Bayesian Hybrid Modeling (GEBHM), is a probabilistic modeling method, based on the Kennedy and O'Hagan framework, that has been continuously scaled-up and industrialized over several years. In this work, we explain the various advancements in GEBHM's methods and demonstrate their impact on several challenging industrial problems.

Steven Atkinson, Sayan Ghosh, Natarajan Chennimalai-Kumar et al.

We describe a method for Bayesian optimization by which one may incorporate data from multiple systems whose quantitative interrelationships are unknown a priori. All general (nonreal-valued) features of the systems are associated with continuous latent variables that enter as inputs into a single metamodel that simultaneously learns the response surfaces of all of the systems. Bayesian inference is used to determine appropriate beliefs regarding the latent variables. We explain how the resulting probabilistic metamodel may be used for Bayesian optimization tasks and demonstrate its implementation on a variety of synthetic and real-world examples, comparing its performance under zero-, one-, and few-shot settings against traditional Bayesian optimization, which usually requires substantially more data from the system of interest.

Sayan Ghosh, Jesper Kristensen, Yiming Zhang et al.

Multi-fidelity Gaussian process is a common approach to address the extensive computationally demanding algorithms such as optimization, calibration and uncertainty quantification. Adaptive sampling for multi-fidelity Gaussian process is a changing task due to the fact that not only we seek to estimate the next sampling location of the design variable, but also the level of the simulator fidelity. This issue is often addressed by including the cost of the simulator as an another factor in the searching criterion in conjunction with the uncertainty reduction metric. In this work, we extent the traditional design of experiment framework for the multi-fidelity Gaussian process by partitioning the prediction uncertainty based on the fidelity level and the associated cost of execution. In addition, we utilize the concept of Believer which quantifies the effect of adding an exploratory design point on the Gaussian process uncertainty prediction. We demonstrated our framework using academic examples as well as a industrial application of steady-state thermodynamic operation point of a fluidized bed process

Suman Kalyan Maity, Abhishek Panigrahi, Sayan Ghosh et al.

In this paper, we develop a content-cum-user based deep learning framework DeepTagRec to recommend appropriate question tags on Stack Overflow. The proposed system learns the content representation from question title and body. Subsequently, the learnt representation from heterogeneous relationship between user and tags is fused with the content representation for the final tag prediction. On a very large-scale dataset comprising half a million question posts, DeepTagRec beats all the baselines; in particular, it significantly outperforms the best performing baseline T agCombine achieving an overall gain of 60.8% and 36.8% in precision@3 and recall@10 respectively. DeepTagRec also achieves 63% and 33.14% maximum improvement in exact-k accuracy and top-k accuracy respectively over TagCombine

Ankan Mullick, Sayan Ghosh, Ritam Dutt et al.

With the increase in online news consumption, to maximize advertisement revenue, news media websites try to attract and retain their readers on their sites. One of the most effective tools for reader engagement is commenting, where news readers post their views as comments against the news articles. Traditionally, it has been assumed that the comments are mostly made against the full article. In this work, we show that present commenting landscape is far from this assumption. Because the readers lack the time to go over an entire article, most of the comments are relevant to only particular sections of an article. In this paper, we build a system which can automatically classify comments against relevant sections of an article. To implement that, we develop a deep neural network based mechanism to find comments relevant to any section and a paragraph wise commenting interface to showcase them. We believe that such a data driven commenting system can help news websites to further increase reader engagement.

Sayan Ghosh, Mathieu Chollet, Eugene Laksana et al.

Human verbal communication includes affective messages which are conveyed through use of emotionally colored words. There has been a lot of research in this direction but the problem of integrating state-of-the-art neural language models with affective information remains an area ripe for exploration. In this paper, we propose an extension to an LSTM (Long Short-Term Memory) language model for generating conversational text, conditioned on affect categories. Our proposed model, Affect-LM enables us to customize the degree of emotional content in generated sentences through an additional design parameter. Perception studies conducted using Amazon Mechanical Turk show that Affect-LM generates naturally looking emotional sentences without sacrificing grammatical correctness. Affect-LM also learns affect-discriminative word representations, and perplexity experiments show that additional affective information in conversational text can improve language model prediction.