arXiv subjects

Niranjan Balasubramanian

Publications and source records attributed to Niranjan Balasubramanian.

At least 19 recordsLinked to original sources

arXiv2026

Break It Down, Pass It On: Cross-Task Skill Transfer in LLM Agents

Large language model (LLM) agents can induce skills from completed tasks and reuse them later to grow more capable with experience. In practice, induced skills may transfer unreliably and can even harm the agent that retrieves them. When agent-induced skills transfer reliably across tasks remains an open question. We conduct a comprehensive and controlled study of how the way skills are induced shapes their transfer across tasks. Specifically, we compare task-level with subtask-level skill induction and text with code skill formats, the two axes along which existing methods differ. Task-level skills mostly reduce the agent's performance below its no-memory baseline while subtask-level skills raise it above on average, and text skills transfer better than code skills. To further understand our findings, we examine two complementary properties of the induced skills: specificity, which measures how closely a skill matches real tasks, and abstractness, which measures how evenly its relevance spreads across tasks. Neither property alone predicts task success, but their combined effect does, which we propose as a skill utility score. The score correlates consistently with task success when skills are transferred, and subtask-level and text skills score higher. Computing skill utility only needs the skills and task descriptions but not any task execution, so our score serves as a practical diagnostic of a skill memory before any new task runs.

cs.AI↗

arXiv2026

Certificates without Electrons? Theory and Evidence on Impacts from AI-Driven Power Demand

Data centers now account for 4.4% of United States electricity demand, yet the grid-level effectiveness of the renewable energy certificates (RECs) and power purchase agreements (PPAs) hyperscalers use to claim carbon neutrality remains unclear. We develop a game-theoretic model in which a data center operator chooses among RECs, PPAs, and behind-the-meter colocation while generators make entry decisions under endogenous financing costs. The model identifies a timing wedge -- the mismatch between consumption and credited renewable generation -- as a central mechanism through which AI demand degrades reliability, raises prices, and increases emissions even when RECs cover 100% of annual consumption. Colocation with storage addresses this wedge directly and induces the greatest renewable entry by eliminating generator revenue risk. We test these predictions by exploiting the staggered release of large language models as a natural experiment, using difference-in-differences on a novel dataset linking AI activity to local grid outcomes. AI demand significantly increases fossil generation, wholesale prices (up to 25% in treated PJM zones), and outage frequency (0.5--1 additional outages per year) near data centers, with impacts scaling in model size. Data centers with on-site generation exhibit a sign reversal in power-quality effects, consistent with the model's prediction that behind-the-meter capacity absorbs demand spikes. Counterfactual analyses show that edge inference, spatial reallocation, and colocated storage each substantially mitigate grid impacts, while REC-only strategies do not. Together, our results demonstrate that the externalities of AI to the grid are tightly coupled to procurement design and the spatial organization of data center infrastructure.

econ.EM↗

arXiv2026

AppWorld-UL: Benchmarking Diverse Agent-User Interactions for Tool-Use

Tool-use agents that address day-to-day digital tasks such as ordering groceries must not only operate applications, but also interact with the user, e.g., to ask clarification questions, prompt for confirmation, and inform the user when the instruction is infeasible. However, current benchmarks for evaluating agent-user interactions do not capture the diversity of such interactions. Further, they operate in small environments with few, often non-state-changing, APIs. To address this gap, we introduce AppWorld-UL, a ``user-in-the-loop'' benchmark of 516 challenging tasks requiring diverse agent-user interactions. Building upon the AppWorld framework with 9 popular simulated apps like Amazon and Spotify, we systematically modify original tasks to introduce ambiguities and constraints that necessitate various types of agent-user interaction. User behavior is simulated by an LLM prompted to respond with carefully designed knowledge boundaries, offering more reliable simulation than the unconstrained or overly rigid alternatives used in prior work. Our evaluation reveals that a state-of-the-art LLM, Claude Opus 4.7, achieves only 48.6% success on AppWorld-UL, and only 35.7% on the harder, compositional subset. On the stricter, scenario-level metric, compositional task performance drops to only 21.3%. Our analysis reveals that correct user-interaction is crucial for success. This demonstrates the benchmark's difficulty and its potential to advance research on user-in-the-loop tool-use agents.

cs.AI↗

arXiv2026

Teaching and Evaluating LLMs to Reason About Polymer Design Related Tasks

Research in AI4Science has shown promise in many science applications, including polymer design. However, current LLMs are ineffective in this problem space because: (i) most models lack polymer-specific knowledge, and (ii) existing aligned models have limited coverage of knowledge and capabilities relevant to polymer design. Addressing this, we introduce PolyBench, a large-scale training and test benchmark dataset of more than 125K polymer design-related tasks, leveraging a knowledge base of more than 13 million data points obtained from experimental and synthetic data sources to ensure broad coverage of polymers and their properties. For effective alignment using PolyBench, we introduce a knowledge-augmented reasoning distillation method that augments this dataset with structured CoT. Furthermore, tasks in PolyBench are organized from simple to complex analytical reasoning problems, enabling generalization tests and diagnostic probes across the problem space. Experiments show that small- and mid- sized language models (SLMs) with 7B to 32BB parameters, trained on PolyBench, outperform similar-sized models and remain competitive with closed-source frontier LLMs on PolyBench's test dataset, while demonstrating performance gains on external polymer benchmarks. Dataset and associated code available at https://github.com/StonyBrookNLP/PolyBench.

cs.CL↗

arXiv2026

Syntax Is Easy, Semantics Is Hard: Evaluating LLMs for LTL Translation

Propositional Linear Temporal Logic (LTL) is a popular formalism for specifying desirable requirements and security and privacy policies for software, networks, and systems. Yet expressing such requirements and policies in LTL remains challenging because of its intricate semantics. Since many security and privacy analysis tools require LTL formulas as input, this difficulty places them out of reach for many developers and analysts. Large Language Models (LLMs) could broaden access to such tools by translating natural language fragments into LTL formulas. This paper evaluates that premise by assessing how effectively several representative LLMs translate assertive English sentences into LTL formulas. Using both human-generated and synthetic ground-truth data, we evaluate effectiveness along syntactic and semantic dimensions. The results reveal three findings: (1) in line with prior findings, LLMs perform better on syntactic aspects of LTL than on semantic ones; (2) they generally benefit from more detailed prompts; and (3) reformulating the task as a Python code-completion problem substantially improves overall performance. We also discuss challenges in conducting a fair evaluation on this task and conclude with recommendations for future work.

cs.LO↗

arXiv2026

Addressing the Ecological Fallacy in Larger LMs with Human Context

Language model training and inference ignore a fundamental linguistic fact -- there is a dependence between multiple sequences of text written by the same person. Prior work has shown that addressing this form of \textit{ecological fallacy} can greatly improve the performance of multiple smaller (~124M) GPT-based models. In this work, we ask if addressing the ecological fallacy by modeling the author's language context with a specific LM task (called HuLM) can provide similar benefits for a larger-scale model, an 8B Llama model. To this end, we explore variants that process an author's language in the context of their other temporally ordered texts. We study the effect of pre-training with this author context using the HuLM objective, as well as using it during fine-tuning with author context (\textit{HuFT:Human-aware Fine-Tuning}). Empirical comparisons show that addressing the ecological fallacy during fine-tuning alone using QLoRA improves the performance of the larger 8B model over standard fine-tuning. Additionally, QLoRA-based continued HuLM pre-training results in a human-aware model generalizable for improved performance over eight downstream tasks with linear task classifier training alone. These results indicate the utility and importance of modeling language in the context of its original generators, the authors.

cs.CL↗

arXiv2025

Fine-Grained Energy Prediction For Parallellized LLM Inference With PIE-P

With the widespread adoption of Large Language Models (LLMs), energy costs of running LLMs is quickly becoming a critical concern. However, precisely measuring the energy consumption of LLMs is often infeasible because hardware-based power monitors are not always accessible and software-based energy measurement tools are not accurate. While various prediction techniques have been developed to estimate LLM energy consumption, these approaches are limited to single-GPU environments and thus are not applicable to modern LLM inference which is typically parallelized across multiple GPUs. In this work, we remedy this gap and introduce PIE-P, a fine-grained energy prediction framework for multi-GPU inference, including tensor, pipeline, and data parallelism. Predicting the energy under parallelized inference is complicated by the non-determinism in inter-GPU communication, additional communication overheads, and difficulties in isolating energy during the communication/synchronization phase. We develop a scalable prediction framework that addresses these issues via precise sampling, fine-grained modeling of inter-GPU communication, and careful accounting of parallelization overhead. Our evaluation results show that PIE-P yields accurate and fine-grained energy predictions across parallelism strategies, significantly outperforming baselines.

cs.DC↗

arXiv2025

Continual Learning with Global Alignment

Continual learning aims to sequentially learn new tasks without forgetting previous tasks' knowledge (catastrophic forgetting). One factor that can cause forgetting is the interference between the gradients on losses from different tasks. When the gradients on the current task's loss are in opposing directions to those on previous tasks' losses, updating the model for the current task may cause performance degradation on previous tasks. In this paper, we first identify causes of the above interference, and hypothesize that correlations between data representations are a key factor of interference. We then propose a method for promoting appropriate correlations between arbitrary tasks' data representations (i.e., global alignment) in individual task learning. Specifically, we learn the data representation as a task-specific composition of pre-trained token representations shared across all tasks. Then the correlations between different tasks' data representations are grounded by correlations between pre-trained token representations. We explore different ways to learn such compositions. Without experience replay, our model achieves SOTA performance in continual learning tasks. It also achieves advanced class-incremental performance through task-incremental training.

cs.CL↗

arXiv2025

ProST: Progressive Sub-task Training for Pareto-Optimal Multi-agent Systems Using Small Language Models

Multi-agent systems with smaller language models (SLMs) present a viable alternative to single agent systems powered by large language models (LLMs) for addressing complex problems. In this work, we study how these alternatives compare in terms of both effectiveness and efficiency. To study this trade-off, we instantiate single and multi-agent systems for the complex problems in the AppWorld environment using different sized language models. We find that difficulties with long-trajectory learning in smaller language models (SLMs) limit their performance. Even when trained for specialized roles, SLMs fail to learn all subtasks effectively. To address this issue, we introduce a simple progressive sub-task training strategy, which introduces new sub-tasks progressively in each training epoch. We find that this novel strategy, analogous to instance level curriculum learning, consistently improves the effectiveness of multi-agents at all configurations. Our Pareto analysis shows that fine-tuned multi-agent systems yield better effectiveness-efficiency trade-offs. Additional ablations and analyses shows the importance of our progressive training strategy and its ability to reduce subtask error rates.

cs.CL↗

arXiv2025

Residualized Similarity for Faithfully Explainable Authorship Verification

Responsible use of Authorship Verification (AV) systems not only requires high accuracy but also interpretable solutions. More importantly, for systems to be used to make decisions with real-world consequences requires the model's prediction to be explainable using interpretable features that can be traced to the original texts. Neural methods achieve high accuracies, but their representations lack direct interpretability. Furthermore, LLM predictions cannot be explained faithfully -- if there is an explanation given for a prediction, it doesn't represent the reasoning process behind the model's prediction. In this paper, we introduce Residualized Similarity (RS), a novel method that supplements systems using interpretable features with a neural network to improve their performance while maintaining interpretability. Authorship verification is fundamentally a similarity task, where the goal is to measure how alike two documents are. The key idea is to use the neural network to predict a similarity residual, i.e. the error in the similarity predicted by the interpretable system. Our evaluation across four datasets shows that not only can we match the performance of state-of-the-art authorship verification models, but we can show how and to what degree the final prediction is faithful and interpretable.

cs.CL↗

arXiv2025

Teaching an Old LLM Secure Coding: Localized Preference Optimization on Distilled Preferences

LLM generated code often contains security issues. We address two key challenges in improving secure code generation. First, obtaining high quality training data covering a broad set of security issues is critical. To address this, we introduce a method for distilling a preference dataset of insecure and secure code pairs from frontier LLMs, along with a security reasoning that explains the issues and the fix. The key idea here is to make use of security knowledge sources to devise a systematic prompting strategy that ensures broad coverage. Second, aligning models to secure code requires focusing on localized regions of code. Direct preference optimization methods, like SimPO, are not designed to handle these localized differences and turn out to be ineffective. We address this with a new localized preference optimization algorithm that masks the security related tokens in both the winning (secure) and losing (insecure) responses. To prevent loss in code quality, we also add a regularizer. Evaluations show that both training on our dataset, DiSCo, and the new preference optimization algorithm, LPO, yield substantial reductions in code insecurity while also improving overall code quality. Code and dataset are available at https://github.com/StonyBrookNLP/disco-lpo.

cs.CR↗

arXiv2025

Handling Open-Vocabulary Constructs in Formalizing Specifications: Retrieval-Augmented Parsing with Expert Knowledge

We study the problem of Open-Vocabulary Constructs(OVCs) -- ones not known beforehand -- in the context of converting natural language (NL) specifications into formal languages (e.g., temporal logic or code). Models fare poorly on OVCs due to a lack of necessary knowledge a priori. In such situations, a domain expert can provide correct constructs at inference time based on their preferences or domain knowledge. Our goal is to effectively reuse this inference-time, expert-provided knowledge for future parses without retraining the model. We present dynamic knowledge-augmented parsing(DKAP), where in addition to the input sentence, the model receives (dynamically growing) expert knowledge as a key-value lexicon that associates NL phrases with correct OVC constructs. We propose ROLex, a retrieval-augmented parsing approach that uses this lexicon. A retriever and a generator are trained to find and use the key-value store to produce the correct parse. A key challenge lies in curating data for this retrieval-augmented parser. We utilize synthetic data generation and the data augmentation techniques on annotated (NL sentence, FL statement) pairs to train the augmented parser. To improve training effectiveness, we propose multiple strategies to teach models to focus on the relevant subset of retrieved knowledge. Finally, we introduce a new evaluation paradigm modeled after the DKAP problem and simulate the scenario across three formalization tasks (NL2LTL, NL2Code, and NL2CMD). Our evaluations show that DKAP is a difficult challenge, and ROLex helps improve the performance of baseline models by using dynamic expert knowledge effectively.

cs.SE↗

arXiv2025

MuSciClaims: Multimodal Scientific Claim Verification

Assessing scientific claims requires identifying, extracting, and reasoning with multimodal data expressed in information-rich figures in scientific literature. Despite the large body of work in scientific QA, figure captioning, and other multimodal reasoning tasks over chart-based data, there are no readily usable multimodal benchmarks that directly test claim verification abilities. To remedy this gap, we introduce a new benchmark MuSciClaims accompanied by diagnostics tasks. We automatically extract supported claims from scientific articles, which we manually perturb to produce contradicted claims. The perturbations are designed to test for a specific set of claim verification capabilities. We also introduce a suite of diagnostic tasks that help understand model failures. Our results show most vision-language models are poor (~0.3-0.5 F1), with even the best model only achieving 0.72 F1. They are also biased towards judging claims as supported, likely misunderstanding nuanced perturbations within the claims. Our diagnostics show models are bad at localizing correct evidence within figures, struggle with aggregating information across modalities, and often fail to understand basic components of the figure.

cs.CL↗

arXiv2025

Causal Graph based Event Reasoning using Semantic Relation Experts

Understanding how events in a scenario causally connect with each other is important for effectively modeling and reasoning about events. But event reasoning remains a difficult challenge, and despite recent advances, Large Language Models (LLMs) still struggle to accurately identify causal connections between events. This struggle leads to poor performance on deeper reasoning tasks like event forecasting and timeline understanding. To address this challenge, we investigate the generation of causal event graphs (e.g., A enables B) as a parallel mechanism to help LLMs explicitly represent causality during inference. This paper evaluates both how to generate correct graphs as well as how graphs can assist reasoning. We propose a collaborative approach to causal graph generation where we use LLMs to simulate experts that focus on specific semantic relations. The experts engage in multiple rounds of discussions which are then consolidated by a final expert. Then, to demonstrate the utility of causal graphs, we use them on multiple downstream applications, and also introduce a new explainable event prediction task that requires a causal chain of events in the explanation. These explanations are more informative and coherent than baseline generations. Finally, our overall approach not finetuned on any downstream task, achieves competitive results with state-of-the-art models on both forecasting and next event prediction tasks.

cs.AI↗

arXiv2025

Quantifying Misattribution Unfairness in Authorship Attribution

Authorship misattribution can have profound consequences in real life. In forensic settings simply being considered as one of the potential authors of an evidential piece of text or communication can result in undesirable scrutiny. This raises a fairness question: Is every author in the candidate pool at equal risk of misattribution? Standard evaluation measures for authorship attribution systems do not explicitly account for this notion of fairness. We introduce a simple measure, Misattribution Unfairness Index (MAUIk), which is based on how often authors are ranked in the top k for texts they did not write. Using this measure we quantify the unfairness of five models on two different datasets. All models exhibit high levels of unfairness with increased risks for some authors. Furthermore, we find that this unfairness relates to how the models embed the authors as vectors in the latent search space. In particular, we observe that the risk of misattribution is higher for authors closer to the centroid (or center) of the embedded authors in the haystack. These results indicate the potential for harm and the need for communicating with and calibrating end users on misattribution risk when building and providing such models for downstream use.

cs.CL↗

arXiv2025

$\texttt{DIAMONDs}$: A Dataset for $\mathbb{D}$ynamic $\mathbb{I}$nformation $\mathbb{A}$nd $\mathbb{M}$ental modeling $\mathbb{O}$f $\mathbb{N}$umeric $\mathbb{D}$iscussions

Understanding multiparty conversations demands robust Theory of Mind (ToM) capabilities, including the ability to track dynamic information, manage knowledge asymmetries, and distinguish relevant information across extended exchanges. To advance ToM evaluation in such settings, we present a carefully designed scalable methodology for generating high-quality benchmark conversation-question pairs with these characteristics. Using this methodology, we create $\texttt{DIAMONDs}$, a new conversational QA dataset covering common business, financial or other group interactions. In these goal-oriented conversations, participants often have to track certain numerical quantities (say $\textit{expected profit}$) of interest that can be derived from other variable quantities (like $\textit{marketing expenses, expected sales, salary}$, etc.), whose values also change over the course of the conversation. $\texttt{DIAMONDs}$ questions pose simple numerical reasoning problems over such quantities of interest (e.g., $\textit{funds required for charity events, expected company profit next quarter}$, etc.) in the context of the information exchanged in conversations. This allows for precisely evaluating ToM capabilities for carefully tracking and reasoning over participants' knowledge states. Our evaluation of state-of-the-art language models reveals significant challenges in handling participant-centric reasoning, specifically in situations where participants have false beliefs. Models also struggle with conversations containing distractors and show limited ability to identify scenarios with insufficient information. These findings highlight current models' ToM limitations in handling real-world multi-party conversations.

cs.AI↗

arXiv2025

Evaluation of LLMs-based Hidden States as Author Representations for Psychological Human-Centered NLP Tasks

Like most of NLP, models for human-centered NLP tasks -- tasks attempting to assess author-level information -- predominantly use representations derived from hidden states of Transformer-based LLMs. However, what component of the LM is used for the representation varies widely. Moreover, there is a need for Human Language Models (HuLMs) that implicitly model the author and provide a user-level hidden state. Here, we systematically evaluate different ways of representing documents and users using different LM and HuLM architectures to predict task outcomes as both dynamically changing states and averaged trait-like user-level attributes of valence, arousal, empathy, and distress. We find that representing documents as an average of the token hidden states performs the best generally. Further, while a user-level hidden state itself is rarely the best representation, we find its inclusion in the model strengthens token or document embeddings used to derive document- and user-level representations resulting in best performances.

cs.CL↗

arXiv2025

CaT-BENCH: Benchmarking Language Model Understanding of Causal and Temporal Dependencies in Plans

Understanding the abilities of LLMs to reason about natural language plans, such as instructional text and recipes, is critical to reliably using them in decision-making systems. A fundamental aspect of plans is the temporal order in which their steps needs to be executed, which reflects the underlying causal dependencies between them. We introduce CaT-Bench, a benchmark of Step Order Prediction questions, which test whether a step must necessarily occur before or after another in cooking recipe plans. We use this to evaluate how well frontier LLMs understand causal and temporal dependencies. We find that SOTA LLMs are underwhelming (best zero-shot is only 0.59 in F1), and are biased towards predicting dependence more often, perhaps relying on temporal order of steps as a heuristic. While prompting for explanations and using few-shot examples improve performance, the best F1 result is only 0.73. Further, human evaluation of explanations along with answer correctness show that, on average, humans do not agree with model reasoning. Surprisingly, we also find that explaining after answering leads to better performance than normal chain-of-thought prompting, and LLM answers are not consistent across questions about the same step pairs. Overall, results show that LLMs' ability to detect dependence between steps has significant room for improvement.

cs.CL↗