Skip to main content
arXiv is now an independent nonprofit! Learn more

Showing 1–17 of 17 results for author: Damani, M

Searching in archive cs. Search in all archives.
.
  1. arXiv:2608.05188  [pdf, ps, other

    cs.CL cs.AI

    Position: It's Time to Optimize LLMs for Self-Consistency

    Authors: Itamar Pres, Belinda Z. Li, Laura Ruis, Zifan Carl Guo, Keya Hu, Mehul Damani, Isha Puri, Ekdeep Singh Lubana, Jacob Andreas

    Abstract: Despite ever-increasing sophistication in language model (LM) pre- and post-training pipelines, many important failures persist: models overcondition on user framing ("sycophancy"), exhibit incomplete logical generalization, and produce confident but incorrect responses. We argue that these failures arise from a modeling assumption permeating all aspects of the pipeline: that behavior can be speci… ▽ More

    Submitted 31 July, 2026; originally announced August 2026.

    Comments: Accepted at the 43rd International Conference on Machine Learning (ICML 2026), Position Paper Track

  2. arXiv:2607.01181  [pdf, ps, other

    cs.LG cs.AI cs.CL

    Right in the Right Way: LM Training with Verifiable Rewards and Human Demonstrations

    Authors: Mehul Damani, Isha Puri, Idan Shenfeld, Jacob Andreas

    Abstract: RL with verifiable rewards (RLVR) has emerged as a powerful paradigm for training LMs on tasks with well-defined success metrics, such as code generation and mathematical reasoning. However, current RLVR methods optimize only what can be objectively scored, often neglecting subjective, non-verifiable aspects of human-like outputs, such as style and structure. This limitation leads to well-document… ▽ More

    Submitted 1 July, 2026; originally announced July 2026.

  3. arXiv:2605.22817  [pdf, ps, other

    cs.LG cs.AI cs.CL cs.NE

    Vector Policy Optimization: Training for Diversity Improves Test-Time Search

    Authors: Ryan Bahlous-Boldi, Isha Puri, Idan Shenfeld, Akarsh Kumar, Mehul Damani, Sebastian Risi, Omar Khattab, Zhang-Wei Hong, Pulkit Agrawal

    Abstract: Language models must now generalize out of the box to novel environments and work inside inference-scaling search procedures, such as AlphaEvolve, that select rollouts with a variety of task-specific reward functions. Unfortunately, the standard paradigm of LLM post-training optimizes a pre-specified scalar reward, often leading current LLMs to produce low-entropy response distributions and thus t… ▽ More

    Submitted 21 May, 2026; originally announced May 2026.

    Comments: 24 pages

  4. arXiv:2603.24844  [pdf, ps, other

    cs.LG cs.AI cs.CL

    Reaching Beyond the Mode: RL for Distributional Reasoning in Language Models

    Authors: Isha Puri, Mehul Damani, Idan Shenfeld, Marzyeh Ghassemi, Jacob Andreas, Yoon Kim

    Abstract: Given a question, a language model (LM) implicitly encodes a distribution over possible answers. In practice, post-training procedures for LMs often collapse this distribution onto a single dominant mode. While this is generally not a problem for benchmark-style evaluations that assume one correct answer, many real-world tasks inherently involve multiple valid answers or irreducible uncertainty. E… ▽ More

    Submitted 25 March, 2026; originally announced March 2026.

  5. arXiv:2603.24366  [pdf, ps, other

    cs.LG cs.RO

    CoordLight: Learning Decentralized Coordination for Network-Wide Traffic Signal Control

    Authors: Yifeng Zhang, Harsh Goel, Peizhuo Li, Mehul Damani, Sandeep Chinchali, Guillaume Sartoretti

    Abstract: Adaptive traffic signal control (ATSC) is crucial in alleviating congestion, maximizing throughput and promoting sustainable mobility in ever-expanding cities. Multi-Agent Reinforcement Learning (MARL) has recently shown significant potential in addressing complex traffic dynamics, but the intricacies of partial observability and coordination in decentralized environments still remain key challeng… ▽ More

    Submitted 25 March, 2026; originally announced March 2026.

    Comments: \c{opyright} 20XX IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

  6. arXiv:2602.24287  [pdf, ps, other

    cs.CL cs.AI

    Do LLMs Benefit From Their Own Words?

    Authors: Jenny Y. Huang, Leshem Choshen, Wei Sun, Omar Khattab, Ramón Fernandez Astudillo, Mehul Damani, Tamara Broderick, Jacob Andreas

    Abstract: In multi-turn conversations, large language models typically condition on the full conversation history: both past user prompts and assistant responses. We revisit this design choice by comparing full-context prompting to four alternative, substantially-reduced context configurations. Analyzing in-the-wild multi-turn conversations across three open reasoning and one state-of-the-art model, we find… ▽ More

    Submitted 11 August, 2026; v1 submitted 27 February, 2026; originally announced February 2026.

  7. arXiv:2601.19897  [pdf, ps, other

    cs.LG

    Self-Distillation Enables Continual Learning

    Authors: Idan Shenfeld, Mehul Damani, Jonas Hübotter, Pulkit Agrawal

    Abstract: Continual learning, enabling models to acquire new skills and knowledge without degrading existing capabilities, remains a fundamental challenge for foundation models. While on-policy reinforcement learning can reduce forgetting, it requires explicit reward functions that are often unavailable. Learning from expert demonstrations, the primary alternative, is dominated by supervised fine-tuning (SF… ▽ More

    Submitted 7 August, 2026; v1 submitted 27 January, 2026; originally announced January 2026.

  8. arXiv:2509.09864  [pdf, ps, other

    cs.LG cs.AI cs.CL

    Latency and Token-Aware Test-Time Compute

    Authors: Jenny Y. Huang, Mehul Damani, Yousef El-Kurdi, Ramon Astudillo, Wei Sun

    Abstract: Inference-time scaling has emerged as a powerful way to improve large language model (LLM) performance by generating multiple candidate responses and selecting among them. However, existing work on dynamic allocation for test-time compute typically considers only parallel generation methods such as best-of-N, overlooking incremental decoding methods like beam search, and has largely ignored latenc… ▽ More

    Submitted 11 September, 2025; originally announced September 2025.

  9. arXiv:2507.16806  [pdf, ps, other

    cs.LG cs.AI cs.CL

    Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty

    Authors: Mehul Damani, Isha Puri, Stewart Slocum, Idan Shenfeld, Leshem Choshen, Yoon Kim, Jacob Andreas

    Abstract: When language models (LMs) are trained via reinforcement learning (RL) to generate natural language "reasoning chains", their performance improves on a variety of difficult question answering tasks. Today, almost all successful applications of RL for reasoning use binary reward functions that evaluate the correctness of LM outputs. Because such reward functions do not penalize guessing or low-conf… ▽ More

    Submitted 14 May, 2026; v1 submitted 22 July, 2025; originally announced July 2025.

  10. arXiv:2411.07279  [pdf, other

    cs.AI cs.CL cs.LG

    The Surprising Effectiveness of Test-Time Training for Few-Shot Learning

    Authors: Ekin Akyürek, Mehul Damani, Adam Zweiger, Linlu Qiu, Han Guo, Jyothish Pari, Yoon Kim, Jacob Andreas

    Abstract: Language models (LMs) have shown impressive performance on tasks within their training distribution, but often struggle with structurally novel tasks even when given a small number of in-context task examples. We investigate the effectiveness of test-time training (TTT) -- temporarily updating model parameters during inference using a loss derived from input data -- as a mechanism for improving LM… ▽ More

    Submitted 24 March, 2025; v1 submitted 11 November, 2024; originally announced November 2024.

    Comments: Preprint

  11. arXiv:2410.04707  [pdf, other

    cs.LG cs.AI cs.CL

    Learning How Hard to Think: Input-Adaptive Allocation of LM Computation

    Authors: Mehul Damani, Idan Shenfeld, Andi Peng, Andreea Bobu, Jacob Andreas

    Abstract: Computationally intensive decoding procedures--including search, reranking, and self-critique--can improve the quality of language model (LM) outputs in problems spanning code generation, numerical reasoning, and dialog. Existing work typically applies the same decoding procedure for every input to an LM. But not all inputs require the same amount of computation to process. Can we allocate decodin… ▽ More

    Submitted 6 October, 2024; originally announced October 2024.

  12. arXiv:2307.15217  [pdf, other

    cs.AI cs.CL cs.LG

    Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback

    Authors: Stephen Casper, Xander Davies, Claudia Shi, Thomas Krendl Gilbert, Jérémy Scheurer, Javier Rando, Rachel Freedman, Tomasz Korbak, David Lindner, Pedro Freire, Tony Wang, Samuel Marks, Charbel-Raphaël Segerie, Micah Carroll, Andi Peng, Phillip Christoffersen, Mehul Damani, Stewart Slocum, Usman Anwar, Anand Siththaranjan, Max Nadeau, Eric J. Michaud, Jacob Pfau, Dmitrii Krasheninnikov, Xin Chen , et al. (7 additional authors not shown)

    Abstract: Reinforcement learning from human feedback (RLHF) is a technique for training AI systems to align with human goals. RLHF has emerged as the central method used to finetune state-of-the-art large language models (LLMs). Despite this popularity, there has been relatively little public work systematizing its flaws. In this paper, we (1) survey open problems and fundamental limitations of RLHF and rel… ▽ More

    Submitted 11 September, 2023; v1 submitted 27 July, 2023; originally announced July 2023.

  13. arXiv:2305.16145  [pdf, other

    cs.LG

    SocialLight: Distributed Cooperation Learning towards Network-Wide Traffic Signal Control

    Authors: Harsh Goel, Yifeng Zhang, Mehul Damani, Guillaume Sartoretti

    Abstract: Many recent works have turned to multi-agent reinforcement learning (MARL) for adaptive traffic signal control to optimize the travel time of vehicles over large urban networks. However, achieving effective and scalable cooperation among junctions (agents) remains an open challenge, as existing methods often rely on extensive, non-generalizable reward shaping or on non-scalable centralized learnin… ▽ More

    Submitted 20 April, 2023; originally announced May 2023.

    Comments: To appear in the International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023)

  14. arXiv:2208.10469  [pdf, other

    cs.AI cs.GT cs.MA econ.TH

    Formal Contracts Mitigate Social Dilemmas in Multi-Agent RL

    Authors: Andreas A. Haupt, Phillip J. K. Christoffersen, Mehul Damani, Dylan Hadfield-Menell

    Abstract: Multi-agent Reinforcement Learning (MARL) is a powerful tool for training autonomous agents acting independently in a common environment. However, it can lead to sub-optimal behavior when individual incentives and group incentives diverge. Humans are remarkably capable at solving these social dilemmas. It is an open problem in MARL to replicate such cooperative behaviors in selfish agents. In this… ▽ More

    Submitted 29 January, 2024; v1 submitted 22 August, 2022; originally announced August 2022.

  15. arXiv:2204.03516  [pdf, other

    cs.RO cs.AI cs.LG cs.MA

    Distributed Reinforcement Learning for Robot Teams: A Review

    Authors: Yutong Wang, Mehul Damani, Pamela Wang, Yuhong Cao, Guillaume Sartoretti

    Abstract: Purpose of review: Recent advances in sensing, actuation, and computation have opened the door to multi-robot systems consisting of hundreds/thousands of robots, with promising applications to automated manufacturing, disaster relief, harvesting, last-mile delivery, port/airport operations, or search and rescue. The community has leveraged model-free multi-agent reinforcement learning (MARL) to de… ▽ More

    Submitted 7 April, 2022; originally announced April 2022.

    Comments: Preprint of the paper submitted to Springer's Current Robotics Reports

  16. arXiv:2103.16511  [pdf, other

    cs.AI cs.LG

    Flatland Competition 2020: MAPF and MARL for Efficient Train Coordination on a Grid World

    Authors: Florian Laurent, Manuel Schneider, Christian Scheller, Jeremy Watson, Jiaoyang Li, Zhe Chen, Yi Zheng, Shao-Hung Chan, Konstantin Makhnev, Oleg Svidchenko, Vladimir Egorov, Dmitry Ivanov, Aleksei Shpilman, Evgenija Spirovska, Oliver Tanevski, Aleksandar Nikov, Ramon Grunder, David Galevski, Jakov Mitrovski, Guillaume Sartoretti, Zhiyao Luo, Mehul Damani, Nilabha Bhattacharya, Shivam Agarwal, Adrian Egli , et al. (2 additional authors not shown)

    Abstract: The Flatland competition aimed at finding novel approaches to solve the vehicle re-scheduling problem (VRSP). The VRSP is concerned with scheduling trips in traffic networks and the re-scheduling of vehicles when disruptions occur, for example the breakdown of a vehicle. While solving the VRSP in various settings has been an active area in operations research (OR) for decades, the ever-growing com… ▽ More

    Submitted 30 March, 2021; originally announced March 2021.

    Comments: 28 pages, 8 figures

  17. PRIMAL2: Pathfinding via Reinforcement and Imitation Multi-Agent Learning -- Lifelong

    Authors: Mehul Damani, Zhiyao Luo, Emerson Wenzel, Guillaume Sartoretti

    Abstract: Multi-agent path finding (MAPF) is an indispensable component of large-scale robot deployments in numerous domains ranging from airport management to warehouse automation. In particular, this work addresses lifelong MAPF (LMAPF) - an online variant of the problem where agents are immediately assigned a new goal upon reaching their current one - in dense and highly structured environments, typical… ▽ More

    Submitted 4 March, 2021; v1 submitted 16 October, 2020; originally announced October 2020.

    Comments: \c{opyright} 20XX IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works