Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning

• URL: http://arxiv.org/abs/2503.07572v1
• Date: Mon, 10 Mar 2025 17:40:43 GMT
• Title: Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning
• Authors: Yuxiao Qu, Matthew Y. R. Yang, Amrith Setlur, Lewis Tunstall, Edward Emanuel Beeching, Ruslan Salakhutdinov, Aviral Kumar,
• Abstract summary: We formalize the problem of optimizing test-time compute as a meta-reinforcement learning (RL) problem.<n>We show that state-of-the-art models do not minimize regret, but one can do so by maximizing a dense reward bonus in conjunction with the outcome 0/1 reward RL.
• Score: 60.67176246634741
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Training models to effectively use test-time compute is crucial for improving the reasoning performance of LLMs. Current methods mostly do so via fine-tuning on search traces or running RL with 0/1 outcome reward, but do these approaches efficiently utilize test-time compute? Would these approaches continue to scale as the budget improves? In this paper, we try to answer these questions. We formalize the problem of optimizing test-time compute as a meta-reinforcement learning (RL) problem, which provides a principled perspective on spending test-time compute. This perspective enables us to view the long output stream from the LLM as consisting of several episodes run at test time and leads us to use a notion of cumulative regret over output tokens as a way to measure the efficacy of test-time compute. Akin to how RL algorithms can best tradeoff exploration and exploitation over training, minimizing cumulative regret would also provide the best balance between exploration and exploitation in the token stream. While we show that state-of-the-art models do not minimize regret, one can do so by maximizing a dense reward bonus in conjunction with the outcome 0/1 reward RL. This bonus is the ''progress'' made by each subsequent block in the output stream, quantified by the change in the likelihood of eventual success. Using these insights, we develop Meta Reinforcement Fine-Tuning, or MRT, a new class of fine-tuning methods for optimizing test-time compute. MRT leads to a 2-3x relative gain in performance and roughly a 1.5x gain in token efficiency for math reasoning compared to outcome-reward RL.

Related papers

• Scaling Test-Time Compute Without Verification or RL is Suboptimal [70.28430200655919]
  We show that finetuning LLMs with verifier-based (VB) methods based on RL or search is far superior to verifier-free (VF) approaches based on distilling or cloning search traces, given a fixed amount of compute/data budget.<n>We corroborate our theory empirically on both didactic and math reasoning problems with 3/8B-sized pre-trained LLMs, where we find verification is crucial for scaling test-time compute.
  arXiv  Detail & Related papers  (2025-02-17T18:43:24Z)
• Rethinking Fine-Tuning when Scaling Test-Time Compute: Limiting Confidence Improves Mathematical Reasoning [32.45574194957491]
  We show that training with cross-entropy loss can be misaligned with pass@N in that pass@N accuracy $it decreases$ with longer training.<n>We suggest a principled, modified training loss that is better aligned to pass@N by limiting model confidence and rescuing pass@N test performance.
  arXiv  Detail & Related papers  (2025-02-11T00:33:31Z)
• On the Emergence of Thinking in LLMs I: Searching for the Right Intuition [34.32871896067864]
  We propose a post-training framework called Reinforcement Learning via Self-Play (RLSP)<n> RLSP involves three steps: supervised fine-tuning with human or synthetic demonstrations of the reasoning process, using an exploration reward signal to encourage diverse and efficient reasoning behaviors, and RL training with an outcome verifier to ensure correctness while preventing reward hacking.<n> Empirical studies in the math domain show that RLSP improves reasoning.
  arXiv  Detail & Related papers  (2025-02-10T18:52:04Z)
• Thinking Before Running! Efficient Code Generation with Thorough Exploration and Optimal Refinement [35.991531332335654]
  We introduce ThinkCoder, a framework that combines thorough exploration with optimal refinement. The exploration phase diversifies the solution space by searching for potential solutions, followed by a refinement phase that enhances precision. This approach allows us to select the best solution through careful consideration before taking action, avoiding excessive trial and error.
  arXiv  Detail & Related papers  (2024-12-30T07:02:15Z)
• Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning [90.23629291067763]
  A promising approach for improving reasoning in large language models is to use process reward models (PRMs) PRMs provide feedback at each step of a multi-step reasoning trace, potentially improving credit assignment over outcome reward models (ORMs) To improve a base policy by running search against a PRM or using it as dense rewards for reinforcement learning (RL), we ask: "How should we design process rewards?" We theoretically characterize the set of good provers and our results show that optimizing process rewards from such provers improves exploration during test-time search and online RL.
  arXiv  Detail & Related papers  (2024-10-10T17:31:23Z)
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters [27.656263126925815]
  We study the scaling of inference-time computation in LLMs. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt.
  arXiv  Detail & Related papers  (2024-08-06T17:35:05Z)
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [55.96599486604344]
  We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process. We use Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data.
  arXiv  Detail & Related papers  (2024-05-01T11:10:24Z)
• Provable and Practical: Efficient Exploration in Reinforcement Learning via Langevin Monte Carlo [104.9535542833054]
  We present a scalable and effective exploration strategy based on Thompson sampling for reinforcement learning (RL) We instead directly sample the Q function from its posterior distribution, by using Langevin Monte Carlo. Our approach achieves better or similar results compared with state-of-the-art deep RL algorithms on several challenging exploration tasks from the Atari57 suite.
  arXiv  Detail & Related papers  (2023-05-29T17:11:28Z)
• Simultaneous Double Q-learning with Conservative Advantage Learning for Actor-Critic Methods [133.85604983925282]
  We propose Simultaneous Double Q-learning with Conservative Advantage Learning (SDQ-CAL) Our algorithm realizes less biased value estimation and achieves state-of-the-art performance in a range of continuous control benchmark tasks.
  arXiv  Detail & Related papers  (2022-05-08T09:17:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.