Automated Theorem Proving (ATP) is a fundamental AI challenge, requiring machine-verifiable proofs in formal languages like Lean to evaluate reasoning. Reinforcement Learning (RL), especially Group Relative Policy Optimization (GRPO), is a mainstream approach. However, GRPO faces two critical issues in ATP: its relativ...
Zhengqing Yan, Fan Guo, Xinyang Liu et al.· 0 citations
Large language models (LLMs) can achieve strong reasoning performance with sufficient computation, but they do not inherently know how much computation a task requires. We study budgeted inference-time reasoning for multiple problems under a strict global token constraint and formulate it as an Ordered Stochastic Multi...
Muyang Zhao, Chunhao Tian, Qi Qi et al.· 0 citations
Merging large language models (LLMs) is a practical way to compose capabilities from multiple fine-tuned checkpoints without retraining. Yet standard schemes (linear weight soups, task vectors, and Fisher-weighted averaging) can preserve loss while quietly destroying alignment. We argue that merging is not a numerical...
Aniruddha Roy, Jyoti Patel, Aman Chadha et al.· 0 citations
Alignment faking is a form of strategic deception in AI in which models selectively comply with training objectives when they infer that they are in training, while preserving different behavior outside training. The phenomenon was first documented for Claude 3 Opus and later examined across additional large language m...
Kartik Garg, Shourya Mishra, Kartikeya Sinha et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Hallucination continues to pose a major obstacle in the reasoning capabilities of large language models (LLMs). Although the Multi-Agent Debate (MAD) paradigm offers a promising solution by promoting consensus among multiple agents to enhance reliability, it relies on the unrealistic assumption that all debaters are ra...
Dayong Liang, Xiao-Yong Wei, Linhui Xiao et al.· 0 citations
Increasing the thinking budget of AI models can significantly improve accuracy, but not all questions warrant the same amount of reasoning. Users may prefer to allocate different amounts of reasoning effort depending on how they value output quality versus latency and cost. To leverage this tradeoff effectively, users...
Michael Kleinman, Matthew Trager, Alessandro Achille et al.· 0 citations
While reinforcement learning agents can achieve superhuman performance in many complex tasks, they typically do not become more computationally efficient as they improve. In contrast, humans gradually require less cognitive effort as they become more proficient at a task. If agents could reason about their compute as t...
Adrian Orenstein, Jiajing Chen, Gwyneth Anne Delos Santos et al.· 0 citations
Chain-of-thought (CoT) reasoning improves LLM accuracy on complex tasks, but CoTs are often unfaithful; that is, the model does not rely on the stated reasoning steps to produce its answer. Existing methods to improve faithfulness require human annotations. We propose LIFT, which automatically scores each reasoning ste...
Anand Swaroop, Maheep Chaudhary, Kevin Zhu et al.· 0 citations
LLM agents are increasingly proposed as substitutes for human participants in social research, usually on the strength of survey responses that match those of the humans they represent. We argue that such output-level alignment is not enough. Drawing on the behavioral sciences, we propose Latent Profiling, a framework...
James Mooney, Josef Woldense, Zheng Robert Jia et al.· 0 citations
Public service systems in many African regions suffer from delayed emergency response and spatial inequity, causing avoidable suffering. This paper introduces OPTIC-ER, a reinforcement learning (RL) framework for real-time, adaptive, and equitable emergency response. OPTIC-ER uses an attention-guided actor-critic archi...
In this paper we try to answer the question "What constitutes Sudoku difficulty rating across different Sudoku websites?" Using two distinct methods that can both solve every Sudoku puzzle, I propose two new metrics to characterize Sudoku difficulty. The first method is based on converting a Sudoku puzzle into its corr...
Large language models can answer questions correctly and still hallucinate their citations. Most citation systems address this by retrieving documents at inference time. Retrieval adds latency and depends on external infrastructure. Retrieval noise can also affect the answer. We study whether LLMs can cite documents fr...
Yukun Huang, Sanxing Chen, Jian Pei et al.· 0 citations
