Neeraj Varshney

Yizhong Wang, Swaroop Mishra, Pegah Alipoormolabashi et al. · allen-ai, amazon-science

How well can NLP models generalize to a variety of unseen tasks when provided with task instructions? To address this question, we first introduce Super-NaturalInstructions, a benchmark of 1,616 diverse NLP tasks and their expert-written instructions. Our collection covers 76 distinct task types, including but not limited to classification, extraction, infilling, sequence tagging, text rewriting, and text composition. This large and diverse collection of tasks enables rigorous benchmarking of cross-task generalization under instructions -- training models to follow instructions on a subset of tasks and evaluating them on the remaining unseen ones. Furthermore, we build Tk-Instruct, a transformer model trained to follow a variety of in-context instructions (plain language task definitions or k-shot examples). Our experiments show that Tk-Instruct outperforms existing instruction-following models such as InstructGPT by over 9% on our benchmark despite being an order of magnitude smaller. We further analyze generalization as a function of various scaling parameters, such as the number of observed tasks, the number of instances per task, and model sizes. We hope our dataset and model facilitate future progress towards more general-purpose NLP models.

Swaroop Mishra, Arindam Mitra, Neeraj Varshney et al. · allen-ai, amazon-science

Given the ubiquitous nature of numbers in text, reasoning with numbers to perform simple calculations is an important skill of AI systems. While many datasets and models have been developed to this end, state-of-the-art AI systems are brittle; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario. Drawing inspiration from GLUE that was proposed in the context of natural language understanding, we propose NumGLUE, a multi-task benchmark that evaluates the performance of AI systems on eight different tasks, that at their core require simple arithmetic understanding. We show that this benchmark is far from being solved with neural models including state-of-the-art large-scale language models performing significantly worse than humans (lower by 46.4%). Further, NumGLUE promotes sharing knowledge across tasks, especially those with limited training data as evidenced by the superior performance (average gain of 3.4% on each task) when a model is jointly trained on all the tasks as opposed to task-specific modeling. Finally, we hope that NumGLUE will encourage systems that perform robust and general arithmetic reasoning within language, a first step towards being able to perform more complex mathematical reasoning.

Neeraj Varshney, Swaroop Mishra, Chitta Baral · amazon-science

In order to equip NLP systems with selective prediction capability, several task-specific approaches have been proposed. However, which approaches work best across tasks or even if they consistently outperform the simplest baseline 'MaxProb' remains to be explored. To this end, we systematically study 'selective prediction' in a large-scale setup of 17 datasets across several NLP tasks. Through comprehensive experiments under in-domain (IID), out-of-domain (OOD), and adversarial (ADV) settings, we show that despite leveraging additional resources (held-out data/computation), none of the existing approaches consistently and considerably outperforms MaxProb in all three settings. Furthermore, their performance does not translate well across tasks. For instance, Monte-Carlo Dropout outperforms all other approaches on Duplicate Detection datasets but does not fare well on NLI datasets, especially in the OOD setting. Thus, we recommend that future selective prediction approaches should be evaluated across tasks and settings for reliable estimation of their capabilities.

Neeraj Varshney, Wenlin Yao, Hongming Zhang et al. · amazon-science

Recently developed large language models have achieved remarkable success in generating fluent and coherent text. However, these models often tend to 'hallucinate' which critically hampers their reliability. In this work, we address this crucial problem and propose an approach that actively detects and mitigates hallucinations during the generation process. Specifically, we first identify the candidates of potential hallucination leveraging the model's logit output values, check their correctness through a validation procedure, mitigate the detected hallucinations, and then continue with the generation process. Through extensive experiments with GPT-3.5 (text-davinci-003) on the 'article generation task', we first demonstrate the individual efficacy of our detection and mitigation techniques. Specifically, the detection technique achieves a recall of ~88% and the mitigation technique successfully mitigates 57.6% of the correctly detected hallucinations. Importantly, our mitigation technique does not introduce new hallucinations even in the case of incorrectly detected hallucinations, i.e., false positives. Then, we show that the proposed active detection and mitigation approach successfully reduces the hallucinations of the GPT-3.5 model from 47.5% to 14.5% on average. We further demonstrate the effectiveness and wide applicability of our approach through additional studies including performance on different types of questions (multi-hop and false premise questions) and with another LLM from a different model family (Vicuna). In summary, our work contributes to improving the reliability and trustworthiness of large language models, a crucial step en route to enabling their widespread adoption in real-world applications.

Man Luo, Shrinidhi Kumbhar, Ming shen et al. · amazon-science

Logical reasoning is fundamental for humans yet presents a substantial challenge in the domain of Artificial Intelligence. Initially, researchers used Knowledge Representation and Reasoning (KR) systems that did not scale and required non-trivial manual effort. Recently, the emergence of large language models (LLMs) has demonstrated the ability to overcome various limitations of formal Knowledge Representation (KR) systems. Consequently, there's a growing interest in using LLMs for logical reasoning via natural language. This work strives to understand the proficiency of LLMs in logical reasoning by offering a brief review of the latest progress in this area; with a focus on the logical reasoning datasets, tasks, and the methods adopted to utilize LLMs for reasoning. To offer a thorough analysis, we have compiled a benchmark titled LogiGLUE. This includes 24 varied datasets encompassing deductive, abductive, and inductive reasoning. Utilizing LogiGLUE as a foundation, we have trained an instruction fine-tuned language model, resulting in LogiT5. We study single-task training, multi-task training, and "chain-of-thought" knowledge distillation fine-tuning technique to assess the performance of model across the different logical reasoning categories. We also assess various LLMs using LogiGLUE, and the findings indicate that LLMs excel most in abductive reasoning, followed by deductive reasoning, while they are least effective at inductive reasoning. We aim to shed light on the capabilities and potential pathways for enhancing logical reasoning proficiency in LLMs, paving the way for more advanced and nuanced developments in this critical field.

Neeraj Varshney, Chitta Baral · amazon-science

Do all instances need inference through the big models for a correct prediction? Perhaps not; some instances are easy and can be answered correctly by even small capacity models. This provides opportunities for improving the computational efficiency of systems. In this work, we present an explorative study on 'model cascading', a simple technique that utilizes a collection of models of varying capacities to accurately yet efficiently output predictions. Through comprehensive experiments in multiple task settings that differ in the number of models available for cascading (K value), we show that cascading improves both the computational efficiency and the prediction accuracy. For instance, in K=3 setting, cascading saves up to 88.93% computation cost and consistently achieves superior prediction accuracy with an improvement of up to 2.18%. We also study the impact of introducing additional models in the cascade and show that it further increases the efficiency improvements. Finally, we hope that our work will facilitate development of efficient NLP systems making their widespread adoption in real-world applications possible.

Neeraj Varshney, Swaroop Mishra, Chitta Baral · amazon-science

Knowledge of questions' difficulty level helps a teacher in several ways, such as estimating students' potential quickly by asking carefully selected questions and improving quality of examination by modifying trivial and hard questions. Can we extract such benefits of instance difficulty in NLP? To this end, we conduct Instance-Level Difficulty Analysis of Evaluation data (ILDAE) in a large-scale setup of 23 datasets and demonstrate its five novel applications: 1) conducting efficient-yet-accurate evaluations with fewer instances saving computational cost and time, 2) improving quality of existing evaluation datasets by repairing erroneous and trivial instances, 3) selecting the best model based on application requirements, 4) analyzing dataset characteristics for guiding future data creation, 5) estimating Out-of-Domain performance reliably. Comprehensive experiments for these applications result in several interesting findings, such as evaluation using just 5% instances (selected via ILDAE) achieves as high as 0.93 Kendall correlation with evaluation using complete dataset and computing weighted accuracy using difficulty scores leads to 5.2% higher correlation with Out-of-Domain performance. We release the difficulty scores and hope our analyses and findings will bring more attention to this important yet understudied field of leveraging instance difficulty in evaluations.

Himanshu Gupta, Neeraj Varshney, Swaroop Mishra et al. · amazon-science

In current NLP research, large-scale language models and their abilities are widely being discussed. Some recent works have also found notable failures of these models. Often these failure examples involve complex reasoning abilities. This work focuses on a simple commonsense ability, reasoning about when an action (or its effect) is feasible. To this end, we introduce FeasibilityQA, a question-answering dataset involving binary classification (BCQ) and multi-choice multi-correct questions (MCQ) that test understanding of feasibility. We show that even state-of-the-art models such as GPT-3, GPT-2, and T5 struggle to answer the feasibility questions correctly. Specifically, on MCQ and BCQ questions, GPT-3 achieves an accuracy of just (19%, 62%) and (25%, 64%) in zero-shot and few-shot settings, respectively. We also evaluate models by providing relevant knowledge statements required to answer the question. We find that the additional knowledge leads to a 7% gain in performance, but the overall performance still remains low. These results make one wonder how much commonsense knowledge about action feasibility is encoded in state-of-the-art models and how well they can reason about it.

Neeraj Varshney, Man Luo, Chitta Baral · amazon-science

Recent state-of-the-art open-domain QA models are typically based on a two stage retriever-reader approach in which the retriever first finds the relevant knowledge/passages and the reader then leverages that to predict the answer. Prior work has shown that the performance of the reader usually tends to improve with the increase in the number of these passages. Thus, state-of-the-art models use a large number of passages (e.g. 100) for inference. While the reader in this approach achieves high prediction performance, its inference is computationally very expensive. We humans, on the other hand, use a more efficient strategy while answering: firstly, if we can confidently answer the question using our already acquired knowledge then we do not even use the external knowledge, and in the case when we do require external knowledge, we don't read the entire knowledge at once, instead, we only read that much knowledge that is sufficient to find the answer. Motivated by this procedure, we ask a research question "Can the open-domain QA reader utilize external knowledge efficiently like humans without sacrificing the prediction performance?" Driven by this question, we explore an approach that utilizes both 'closed-book' (leveraging knowledge already present in the model parameters) and 'open-book' inference (leveraging external knowledge). Furthermore, instead of using a large fixed number of passages for open-book inference, we dynamically read the external knowledge in multiple 'knowledge iterations'. Through comprehensive experiments on NQ and TriviaQA datasets, we demonstrate that this dynamic reading approach improves both the 'inference efficiency' and the 'prediction accuracy' of the reader. Comparing with the FiD reader, this approach matches its accuracy by utilizing just 18.32% of its reader inference cost and also outperforms it by achieving up to 55.10% accuracy on NQ Open.

Neeraj Varshney, Swaroop Mishra, Chitta Baral · amazon-science

Curriculum learning strategies in prior multi-task learning approaches arrange datasets in a difficulty hierarchy either based on human perception or by exhaustively searching the optimal arrangement. However, human perception of difficulty may not always correlate well with machine interpretation leading to poor performance and exhaustive search is computationally expensive. Addressing these concerns, we propose two classes of techniques to arrange training instances into a learning curriculum based on difficulty scores computed via model-based approaches. The two classes i.e Dataset-level and Instance-level differ in granularity of arrangement. Through comprehensive experiments with 12 datasets, we show that instance-level and dataset-level techniques result in strong representations as they lead to an average performance improvement of 4.17% and 3.15% over their respective baselines. Furthermore, we find that most of this improvement comes from correctly answering the difficult instances, implying a greater efficacy of our techniques on difficult tasks.

Ayushi Agarwal, Nisarg Patel, Neeraj Varshney et al. · amazon-science

Though state-of-the-art (SOTA) NLP systems have achieved remarkable performance on a variety of language understanding tasks, they primarily focus on questions that have a correct and a definitive answer. However, in real-world applications, users often ask questions that don't have a definitive answer. Incorrectly answering such questions certainly hampers a system's reliability and trustworthiness. Can SOTA models accurately identify such questions and provide a reasonable response? To investigate the above question, we introduce QnotA, a dataset consisting of five different categories of questions that don't have definitive answers. Furthermore, for each QnotA instance, we also provide a corresponding QA instance i.e. an alternate question that ''can be'' answered. With this data, we formulate three evaluation tasks that test a system's ability to 'identify', 'distinguish', and 'justify' QnotA questions. Through comprehensive experiments, we show that even SOTA models including GPT-3 and Flan T5 do not fare well on these tasks and lack considerably behind the human performance baseline. We conduct a thorough analysis which further leads to several interesting findings. Overall, we believe our work and findings will encourage and facilitate further research in this important area and help develop more robust models.

Chaitanya Dwivedi, Binxuan Huang, Himanshu Gupta et al. · amazon-science

Mixture-of-Experts (MoE) has become the dominant architecture for scaling large language models: frontier models routinely decouple total parameters from per-token computation through sparse expert routing. Scaling laws show that under fixed active computation, model quality scales predictably with total parameters, and MoEs realize this by increasing expert count. However, training large MoEs is expensive, as memory requirements and inter-device communication both scale with total parameter count. We propose expert upcycling, a method for progressively expanding MoE capacity by increasing the number of experts during continued pre-training (CPT). Given a trained E-expert model, the upcycling operator constructs an mE-expert model through expert duplication and router extension while holding top-K routing fixed, preserving per-token inference cost. Duplication provides a warm initialization: the expanded model inherits the source checkpoint's learned representations, starting from a substantially lower loss than random initialization. Subsequent CPT then breaks the symmetry among duplicated experts to drive specialization. We formalize the upcycling operator and develop a theoretical framework decomposing the quality gap into a capacity term and an initialization term. We further introduce utility-based expert selection, which uses gradient-based importance scores to guide non-uniform duplication, more than tripling gap closure when CPT is limited. In our 7B-13B total parameter experiments, the upcycled model matches the fixed-size baseline on validation loss while saving 32% of GPU hours. Comprehensive ablations across model scales, activation ratios, MoE architectures, and training budgets yield a practical recipe for deploying expert upcycling, establishing it as a principled, compute-efficient alternative to training large MoE models from scratch.

Neeraj Varshney, Agneet Chatterjee, Mihir Parmar et al. · amazon-science

Large Language Models (LLMs) have achieved remarkable performance across a wide variety of natural language tasks; however, their large size makes their inference slow and computationally expensive. Focusing on this problem, we propose to instruction tune LLMs with additional explicit losses from the intermediate layers (LITE) and show that it enables these layers to acquire 'good' generation ability without affecting the generation ability of the final layer. We perform 'dynamic confidence-based early exiting' at token level from the intermediate layers which improves the efficiency of text generation without compromising the quality of the generation. We conduct comprehensive experiments by instruction tuning LLaMA-2 models on the Alpaca dataset and holistically evaluate on four different human-instruction test sets. We show that dynamic early exiting achieves consistent and considerable inference computation cost improvements (37.86% for 7B and 46.35% for 13B model) while maintaining the generation quality of the responses. We further conduct a thorough analysis of the results over several important aspects, such as comparing the semantic similarity of the outputs and dissecting the efficiency improvements by comparing the number of tokens generated in the output. In summary, our work contributes to improving the efficiency of LLM inference while maintaining the generation quality, a crucial step en route to enabling their widespread adoption.

Tung Thai, Ming Shen, Mayank Garg et al. · amazon-science

Learning to detect, characterize and accommodate novelties is a challenge that agents operating in open-world domains need to address to be able to guarantee satisfactory task performance. Certain novelties (e.g., changes in environment dynamics) can interfere with the performance or prevent agents from accomplishing task goals altogether. In this paper, we introduce general methods and architectural mechanisms for detecting and characterizing different types of novelties, and for building an appropriate adaptive model to accommodate them utilizing logical representations and reasoning methods. We demonstrate the effectiveness of the proposed methods in evaluations performed by a third party in the adversarial multi-agent board game Monopoly. The results show high novelty detection and accommodation rates across a variety of novelty types, including changes to the rules of the game, as well as changes to the agent's action capabilities.

Himanshu Gupta, Pratik Jayarao, Chaitanya Dwivedi et al.

Code-mixing and code-switching (CSW) remain challenging phenomena for large language models (LLMs). Despite recent advances in multilingual modeling, LLMs often struggle in mixed-language settings, exhibiting systematic degradation in grammaticality, factuality, and safety behavior. This work provides a comprehensive overview of CSW research in modern large language model settings. We introduce a unifying taxonomy that organizes prior work along dimensions of data, modeling, and evaluation, and we distill these findings into a practical playbook of actionable recommendations for building, adapting, and evaluating CSW-capable LLMs. We review modeling approaches ranging from CSW-tailored pre-training and task-specific post-training to prompting strategies and in-context learning. We analyze current evaluation practices, highlighting sources of instability and limited reproducibility, and we catalog existing benchmarks while critically examining their linguistic coverage and English-centric biases. Finally, we discuss emerging safety concerns, including use of code-mixing as a mechanism for bypassing model safeguards, and identify open research challenges.

Mihir Parmar, Nisarg Patel, Neeraj Varshney et al. · amazon-science

Recently developed large language models (LLMs) have been shown to perform remarkably well on a wide range of language understanding tasks. But, can they really "reason" over the natural language? This question has been receiving significant research attention and many reasoning skills such as commonsense, numerical, and qualitative have been studied. However, the crucial skill pertaining to 'logical reasoning' has remained underexplored. Existing work investigating this reasoning ability of LLMs has focused only on a couple of inference rules (such as modus ponens and modus tollens) of propositional and first-order logic. Addressing the above limitation, we comprehensively evaluate the logical reasoning ability of LLMs on 25 different reasoning patterns spanning over propositional, first-order, and non-monotonic logics. To enable systematic evaluation, we introduce LogicBench, a natural language question-answering dataset focusing on the use of a single inference rule. We conduct detailed analysis with a range of LLMs such as GPT-4, ChatGPT, Gemini, Llama-2, and Mistral using chain-of-thought prompting. Experimental results show that existing LLMs do not fare well on LogicBench; especially, they struggle with instances involving complex reasoning and negations. Furthermore, they sometimes overlook contextual information necessary for reasoning to arrive at the correct conclusion. We believe that our work and findings facilitate future research for evaluating and enhancing the logical reasoning ability of LLMs. Data and code are available at https://github.com/Mihir3009/LogicBench.

Jiarui Liu, Kaustubh Dhole, Yingheng Wang et al.

Reinforcement learning with verifiable rewards (RLVR) has recently emerged as a promising framework for aligning language models with complex reasoning objectives. However, most existing methods optimize only for final task outcomes, leaving models vulnerable to collapse when negative rewards dominate early training. This challenge is especially pronounced in honesty alignment, where models must not only solve answerable queries but also identify when conclusions cannot be drawn from the given premises. Deductive reasoning provides an ideal testbed because it isolates reasoning capability from reliance on external factual knowledge. To investigate honesty alignment, we curate two multi-step deductive reasoning datasets from graph structures, one for linear algebra and one for logical inference, and introduce unanswerable cases by randomly perturbing an edge in half of the instances. We find that GRPO, with or without supervised fine tuning initialization, struggles on these tasks. Through extensive experiments across three models, we evaluate stabilization strategies and show that curriculum learning provides some benefit but requires carefully designed in distribution datasets with controllable difficulty. To address these limitations, we propose Anchor, a reinforcement learning method that injects ground truth trajectories into rollouts, preventing early training collapse. Our results demonstrate that this method stabilizes learning and significantly improves the overall reasoning performance, underscoring the importance of training dynamics for enabling reliable deductive reasoning in aligned language models.

Pratik Jayarao, Himanshu Gupta, Neeraj Varshney et al. · amazon-science

As Large Language Models (LLMs) are increasingly adopted as automated judges in benchmarking and reward modeling, ensuring their reliability, efficiency, and robustness has become critical. In this work, we present a systematic comparison of "thinking" and "non-thinking" LLMs in the LLM-as-a-judge paradigm using open-source Qwen 3 models of relatively small sizes (0.6B, 1.7B, and 4B parameters). We evaluate both accuracy and computational efficiency (FLOPs) on RewardBench tasks, and further examine augmentation strategies for non-thinking models, including in-context learning, rubric-guided judging, reference-based evaluation, and n-best aggregation. Our results show that despite these enhancements, non-thinking models generally fall short of their thinking counterparts. Our results show that thinking models achieve approximately 10% points higher accuracy with little overhead (under 2x), in contrast to augmentation strategies like few-shot learning, which deliver modest gains at a higher cost (>8x). Bias and robustness analyses further demonstrate that thinking models maintain significantly greater consistency under a variety of bias conditions such as positional, bandwagon, identity, diversity, and random biases (6% higher on average). We further extend our experiments to the multilingual setting and our results confirm that explicit reasoning extends its benefits beyond English. Overall, our work results in several important findings that provide systematic evidence that explicit reasoning offers clear advantages in the LLM-as-a-judge paradigm not only in accuracy and efficiency but also in robustness.

Nisarg Patel, Mohith Kulkarni, Mihir Parmar et al.

As Large Language Models (LLMs) continue to exhibit remarkable performance in natural language understanding tasks, there is a crucial need to measure their ability for human-like multi-step logical reasoning. Existing logical reasoning evaluation benchmarks often focus primarily on simplistic single-step or multi-step reasoning with a limited set of inference rules. Furthermore, the lack of datasets for evaluating non-monotonic reasoning represents a crucial gap since it aligns more closely with human-like reasoning. To address these limitations, we propose Multi-LogiEval, a comprehensive evaluation dataset encompassing multi-step logical reasoning with various inference rules and depths. Multi-LogiEval covers three logic types--propositional, first-order, and non-monotonic--consisting of more than 30 inference rules and more than 60 of their combinations with various depths. Leveraging this dataset, we conduct evaluations on a range of LLMs including GPT-4, ChatGPT, Gemini-Pro, Yi, Orca, and Mistral, employing a zero-shot chain-of-thought. Experimental results show that there is a significant drop in the performance of LLMs as the reasoning steps/depth increases (average accuracy of ~68% at depth-1 to ~43% at depth-5). We further conduct a thorough investigation of reasoning chains generated by LLMs which reveals several important findings. We believe that Multi-LogiEval facilitates future research for evaluating and enhancing the logical reasoning ability of LLMs. Data is available at https://github.com/Mihir3009/Multi-LogiEval.

Neeraj Varshney, Satyam Raj, Venkatesh Mishra et al.

Large Language Models (LLMs) have achieved remarkable performance across a wide variety of natural language tasks. However, they have been shown to suffer from a critical limitation pertinent to 'hallucination' in their output. Recent research has focused on investigating and addressing this problem for a variety of tasks such as biography generation, question answering, abstractive summarization, and dialogue generation. However, the crucial aspect pertaining to 'negation' has remained considerably underexplored. Negation is important because it adds depth and nuance to the understanding of language and is also crucial for logical reasoning and inference. In this work, we address the above limitation and particularly focus on studying the impact of negation in LLM hallucinations. Specifically, we study four tasks with negation: 'false premise completion', 'constrained fact generation', 'multiple choice question answering', and 'fact generation'. We show that open-source state-of-the-art LLMs such as LLaMA-2-chat, Vicuna, and Orca-2 hallucinate considerably on all these tasks involving negation which underlines a critical shortcoming of these models. Addressing this problem, we further study numerous strategies to mitigate these hallucinations and demonstrate their impact.

Neeraj Varshney, Mihir Parmar, Nisarg Patel et al.

Pre-training on large corpora of text enables the language models to acquire a vast amount of factual and commonsense knowledge which allows them to achieve remarkable performance on a variety of language understanding tasks. They typically acquire this knowledge by learning from the pre-training text and capturing certain patterns from it. However, real-world settings often present scenarios that do not abide by these patterns i.e. scenarios that break the common assumptions. Can state-of-the-art NLP models correctly reason over the contexts of such scenarios? Addressing the above question, in this paper, we investigate the ability of models to correctly reason over contexts that break the common assumptions. To this end, we first systematically create evaluation data in which each data instance consists of (a) a common assumption, (b) a context that follows the assumption, (c) a context that breaks the assumption, and (d) questions based on the contexts. Then, through evaluations on multiple models including GPT-3 and Flan T5, we show that while doing fairly well on contexts that follow the common assumptions, the models struggle to correctly reason over contexts that break those assumptions. Specifically, the performance gap is as high as 20% absolute points. Furthermore, we thoroughly analyze these results revealing several interesting findings. We believe our work and findings will encourage and facilitate further research in developing more robust models that can also reliably reason over contexts that break the common assumptions. Data is available at \url{https://github.com/nrjvarshney/break_the_common_assumptions}.

Neeraj Varshney, Pavel Dolin, Agastya Seth et al. · amazon-science

As Large Language Models (LLMs) play an increasingly pivotal role in natural language processing applications, their safety concerns become critical areas of NLP research. This paper presents Safety and Over-Defensiveness Evaluation (SODE) benchmark: a collection of diverse safe and unsafe prompts with carefully designed evaluation methods that facilitate systematic evaluation, comparison, and analysis over 'safety' and 'over-defensiveness.' With SODE, we study a variety of LLM defense strategies over multiple state-of-the-art LLMs, which reveals several interesting and important findings, such as (a) the widely popular 'self-checking' techniques indeed improve the safety against unsafe inputs, but this comes at the cost of extreme over-defensiveness on the safe inputs, (b) providing a safety instruction along with in-context exemplars (of both safe and unsafe inputs) consistently improves safety and also mitigates undue over-defensiveness of the models, (c) providing contextual knowledge easily breaks the safety guardrails and makes the models more vulnerable to generating unsafe responses. Overall, our work reveals numerous such critical findings that we believe will pave the way and facilitate further research in improving the safety of LLMs.

Neeraj Varshney, Steve Blandino, Jian Wang et al.

ISAC is currently being standardized within the 3GPP New Radio (NR) to enable cellular infrastructure to perform sensing using existing communication waveforms. While standardization is progressing, practical deployment may be limited by scenario-dependent observability constraints. For example, in UMa-AV scenarios, sensing with a single TRP can be affected by restricted angular coverage, partial blockage, and limited field of view, which may degrade detection reliability in three-dimensional UAV environments. For this reason, multi-TRP solutions have been suggested to improve spatial diversity and sensing robustness. In this paper, we present a system-level investigation of multi-TRP assisted monostatic sensing for UAV detection under standardized 3GPP UMa-AV channel assumptions and Release 19 evaluation parameters. We propose a spatial diversity fusion framework and evaluate the achievable performance of a 3GPP network by combining the measurements obtained independently at different TRP. Extensive evaluations demonstrate that multi-TRP assistance improves target observability, reduces spurious detections, and tightens localization error distributions at the cost of additional sensing overhead due to the need for multiple TRPs to periodically allocate radio resources for sensing measurements. In the evaluated scenario, results show that a voting threshold of two assisting TRPs achieves an optimal trade-off between miss detection probability and false alarm suppression, meeting 3GPP performance objectives. Furthermore, we quantify the sensing overhead and show that proper system design, tuned to the application requirements, can substantially reduce its impact: for example, extending the sensing refresh interval beyond the 128 ms coherent processing interval to 1 s reduces the effective overhead from 29 % to approximately 3.7 %, enabling more scalable network deployment.

Steve Blandino, Neeraj Varshney, Jian Wang et al.

3GPP Release 19 has initiated the standardization of integrated sensing and communications (ISAC), including a channel model for monostatic sensing, evaluation scenarios, and performance assessment methodologies. These common assumptions provide an important basis for ISAC evaluation, but reproducible end-to-end studies still require a transparent sensing implementation. This paper evaluates 5G New Radio (NR) base station (gNB)-based monostatic sensing for the Unmanned Aerial Vehicle (UAV) use case using a 5G NR downlink Cyclic Prefix-Orthogonal Frequency Division Multiplexing (CP-OFDM) waveform and positioning reference signals (PRS), following 3GPP Urban Macro-Aerial Vehicle (UMa-AV) scenario assumptions. We present an end-to-end processing chain for multi-target detection and 3D localization, achieving more than 70% detection probability with less than 5% false alarm rate, in the considered scenario. For correctly detected targets, localization errors are on the order of a few meters, with a 90th-percentile error of 4m and 6m in the vertical and horizontal directions, respectively. To support reproducible baseline studies and further research, we release the simulator 5GNRad, which reproduces our evaluation

Aswin RRV, Nemika Tyagi, Md Nayem Uddin et al.

This study explores the sycophantic tendencies of Large Language Models (LLMs), where these models tend to provide answers that match what users want to hear, even if they are not entirely correct. The motivation behind this exploration stems from the common behavior observed in individuals searching the internet for facts with partial or misleading knowledge. Similar to using web search engines, users may recall fragments of misleading keywords and submit them to an LLM, hoping for a comprehensive response. Our empirical analysis of several LLMs shows the potential danger of these models amplifying misinformation when presented with misleading keywords. Additionally, we thoroughly assess four existing hallucination mitigation strategies to reduce LLMs sycophantic behavior. Our experiments demonstrate the effectiveness of these strategies for generating factually correct statements. Furthermore, our analyses delve into knowledge-probing experiments on factual keywords and different categories of sycophancy mitigation.

Neeraj Varshney, Himanshu Gupta, Eric Robertson et al.

State-of-the-art natural language processing models have been shown to achieve remarkable performance in 'closed-world' settings where all the labels in the evaluation set are known at training time. However, in real-world settings, 'novel' instances that do not belong to any known class are often observed. This renders the ability to deal with novelties crucial. To initiate a systematic research in this important area of 'dealing with novelties', we introduce 'NoveltyTask', a multi-stage task to evaluate a system's performance on pipelined novelty 'detection' and 'accommodation' tasks. We provide mathematical formulation of NoveltyTask and instantiate it with the authorship attribution task that pertains to identifying the correct author of a given text. We use Amazon reviews corpus and compile a large dataset (consisting of 250k instances across 200 authors/labels) for NoveltyTask. We conduct comprehensive experiments and explore several baseline methods for the task. Our results show that the methods achieve considerably low performance making the task challenging and leaving sufficient room for improvement. Finally, we believe our work will encourage research in this underexplored area of dealing with novelties, an important step en route to developing robust systems.

Neeraj Varshney, Chitta Baral

Despite remarkable progress made in natural language processing, even the state-of-the-art models often make incorrect predictions. Such predictions hamper the reliability of systems and limit their widespread adoption in real-world applications. 'Selective prediction' partly addresses the above concern by enabling models to abstain from answering when their predictions are likely to be incorrect. While selective prediction is advantageous, it leaves us with a pertinent question 'what to do after abstention'. To this end, we present an explorative study on 'Post-Abstention', a task that allows re-attempting the abstained instances with the aim of increasing 'coverage' of the system without significantly sacrificing its 'accuracy'. We first provide mathematical formulation of this task and then explore several methods to solve it. Comprehensive experiments on 11 QA datasets show that these methods lead to considerable risk improvements -- performance metric of the Post-Abstention task -- both in the in-domain and the out-of-domain settings. We also conduct a thorough analysis of these results which further leads to several interesting findings. Finally, we believe that our work will encourage and facilitate further research in this important area of addressing the reliability of NLP systems.

Neeraj Varshney, Pratyay Banerjee, Tejas Gokhale et al.

Transformer-based models achieve impressive performance on numerous Natural Language Inference (NLI) benchmarks when trained on respective training datasets. However, in certain cases, training samples may not be available or collecting them could be time-consuming and resource-intensive. In this work, we address the above challenge and present an explorative study on unsupervised NLI, a paradigm in which no human-annotated training samples are available. We investigate it under three settings: PH, P, and NPH that differ in the extent of unlabeled data available for learning. As a solution, we propose a procedural data generation approach that leverages a set of sentence transformations to collect PHL (Premise, Hypothesis, Label) triplets for training NLI models, bypassing the need for human-annotated training data. Comprehensive experiments with several NLI datasets show that the proposed approach results in accuracies of up to 66.75%, 65.9%, 65.39% in PH, P, and NPH settings respectively, outperforming all existing unsupervised baselines. Furthermore, fine-tuning our model with as little as ~0.1% of the human-annotated training dataset (500 instances) leads to 12.2% higher accuracy than the model trained from scratch on the same 500 instances. Supported by this superior performance, we conclude with a recommendation for collecting high-quality task-specific data.

Neeraj Varshney, Swaroop Mishra, Chitta Baral

Standard NLP tasks do not incorporate several common real-world scenarios such as seeking clarifications about the question, taking advantage of clues, abstaining in order to avoid incorrect answers, etc. This difference in task formulation hinders the adoption of NLP systems in real-world settings. In this work, we take a step towards bridging this gap and present a multi-stage task that simulates a typical human-human questioner-responder interaction such as an interview. Specifically, the system is provided with question simplifications, knowledge statements, examples, etc. at various stages to improve its prediction when it is not sufficiently confident. We instantiate the proposed task in Natural Language Inference setting where a system is evaluated on both in-domain and out-of-domain (OOD) inputs. We conduct comprehensive experiments and find that the multi-stage formulation of our task leads to OOD generalization performance improvement up to 2.29% in Stage 1, 1.91% in Stage 2, 54.88% in Stage 3, and 72.02% in Stage 4 over the standard unguided prediction. However, our task leaves a significant challenge for NLP researchers to further improve OOD performance at each stage.

Pratyay Banerjee, Chitta Baral, Man Luo et al.

A recent work has shown that transformers are able to "reason" with facts and rules in a limited setting where the rules are natural language expressions of conjunctions of conditions implying a conclusion. Since this suggests that transformers may be used for reasoning with knowledge given in natural language, we do a rigorous evaluation of this with respect to a common form of knowledge and its corresponding reasoning -- the reasoning about effects of actions. Reasoning about action and change has been a top focus in the knowledge representation subfield of AI from the early days of AI and more recently it has been a highlight aspect in common sense question answering. We consider four action domains (Blocks World, Logistics, Dock-Worker-Robots and a Generic Domain) in natural language and create QA datasets that involve reasoning about the effects of actions in these domains. We investigate the ability of transformers to (a) learn to reason in these domains and (b) transfer that learning from the generic domains to the other domains.

Neeraj Varshney, Swaroop Mishra, Chitta Baral

It's better to say "I can't answer" than to answer incorrectly. This selective prediction ability is crucial for NLP systems to be reliably deployed in real-world applications. Prior work has shown that existing selective prediction techniques fail to perform well, especially in the out-of-domain setting. In this work, we propose a method that improves probability estimates of models by calibrating them using prediction confidence and difficulty score of instances. Using these two signals, we first annotate held-out instances and then train a calibrator to predict the likelihood of correctness of the model's prediction. We instantiate our method with Natural Language Inference (NLI) and Duplicate Detection (DD) tasks and evaluate it in both In-Domain (IID) and Out-of-Domain (OOD) settings. In (IID, OOD) settings, we show that the representations learned by our calibrator result in an improvement of (15.81%, 5.64%) and (6.19%, 13.9%) over 'MaxProb' -- a selective prediction baseline -- on NLI and DD tasks respectively.

Swaroop Mishra, Arindam Mitra, Neeraj Varshney et al.

Numerical reasoning is often important to accurately understand the world. Recently, several format-specific datasets have been proposed, such as numerical reasoning in the settings of Natural Language Inference (NLI), Reading Comprehension (RC), and Question Answering (QA). Several format-specific models and architectures in response to those datasets have also been proposed. However, there exists a strong need for a benchmark which can evaluate the abilities of models, in performing question format independent numerical reasoning, as (i) the numerical reasoning capabilities we want to teach are not controlled by question formats, (ii) for numerical reasoning technology to have the best possible application, it must be able to process language and reason in a way that is not exclusive to a single format, task, dataset or domain. In pursuit of this goal, we introduce NUMBERGAME, a multifaceted benchmark to evaluate model performance across numerical reasoning tasks of eight diverse formats. We add four existing question types in our compilation. Two of the new types we add are about questions that require external numerical knowledge, commonsense knowledge and domain knowledge. For building a more practical numerical reasoning system, NUMBERGAME demands four capabilities beyond numerical reasoning: (i) detecting question format directly from data (ii) finding intermediate common format to which every format can be converted (iii) incorporating commonsense knowledge (iv) handling data imbalance across formats. We build several baselines, including a new model based on knowledge hunting using a cheatsheet. However, all baselines perform poorly in contrast to the human baselines, indicating the hardness of our benchmark. Our work takes forward the recent progress in generic system development, demonstrating the scope of these under-explored tasks.