Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning
Reviewed by Pithpith:VL5O74ICopen to challenge →
read the original abstract
We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process inspired by the successful strategy employed by AlphaZero. Our work leverages 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. To enhance consistency in intermediate steps, we combine outcome validation and stepwise self-evaluation, continually updating the quality assessment of newly generated data. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data. Theoretical analysis reveals the importance of using on-policy sampled data for successful self-improving. Extensive evaluations on various arithmetic and commonsense reasoning tasks demonstrate remarkable performance improvements over existing models. For instance, our approach outperforms the Mistral-7B Supervised Fine-Tuning (SFT) baseline on GSM8K, MATH, and ARC-C, with substantial increases in accuracy to $81.8\%$ (+$5.9\%$), $34.7\%$ (+$5.8\%$), and $76.4\%$ (+$15.8\%$), respectively. Additionally, our research delves into the training and inference compute tradeoff, providing insights into how our method effectively maximizes performance gains. Our code is publicly available at https://github.com/YuxiXie/MCTS-DPO.
This paper has not been read by Pith yet.
Forward citations
Cited by 36 Pith papers
-
Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing
Local Branch Routing (LBR) is a token-level framework for test-time scaling in language models that uses local branch hidden states for routing and supports end-to-end RL, showing gains in Pass@1 and Pass@32 on math r...
-
Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing
LBR performs token-level test-time scaling via local branch routing on hidden states, enabling end-to-end RL training and improving Pass@1 and Pass@32 on math benchmarks over CoT and RLVR baselines.
-
ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning
ReSum trains LLMs via RLVR to self-summarize reasoning trajectories, yielding 4% average performance gains and 18.6% shorter rollouts through contrastive rollout branches.
-
Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces
Introduces OPT* tasks and two training regimes (solver-guided online policy optimization with rank-based reward shaping and search-based offline RL) plus a theoretical link between search success and information extra...
-
ATLAS: Agentic Test-time Learning-to-Allocate Scaling
ATLAS introduces an LLM-orchestrated agentic framework for dynamic test-time scaling via extensible 'explore' actions, achieving higher accuracy with fewer API calls than fixed-workflow baselines on four benchmarks.
-
CA-SQL: Complexity-Aware Inference Time Reasoning for Text-to-SQL via Exploration and Compute Budget Allocation
CA-SQL achieves 51.72% execution accuracy on the challenging tier of the BIRD benchmark using GPT-4o-mini by scaling exploration breadth according to estimated task difficulty, evolutionary prompt seeding, and candida...
-
StoryAlign: Evaluating and Training Reward Models for Story Generation
StoryReward, trained on a new 100k story preference dataset, sets state-of-the-art performance on the introduced StoryRMB benchmark for aligning LLM stories with human preferences.
-
The Art of Scaling Reinforcement Learning Compute for LLMs
A 400k+ GPU-hour study shows RL scaling in LLMs follows predictable sigmoidal trajectories, with most design choices affecting efficiency rather than the performance asymptote, enabling accurate large-scale prediction...
-
Addressing Over-Refusal in LLMs with Competing Rewards
SEAR trains one LLM via adversarial process rewards to explore harmful reasoning paths but flip to safe outputs, reducing over-refusal while preserving safety.
-
APPO: Agentic Procedural Policy Optimization
APPO refines branching and credit assignment in agentic RL via a Branching Score and procedure-level scaling, improving baselines by nearly 4 points on 13 benchmarks.
-
TRACE: A Unified Rollout Budget Allocation Framework for Efficient Agentic Reinforcement Learning
TRACE is a rollout budget allocation framework that models ReAct turns as tree nodes and uses a predictor to allocate samples to informative prefixes, yielding a 2.8-point accuracy gain on Multi-Hop QA at equal cost.
-
Beyond Fixed Budgets: Characterizing the Inelasticity and Limitations of Tree-of-Thought Reasoning Strategies
DPTS shows cold-start bottlenecks at low budgets while SSDP exhibits frontier depletion, indicating fixed ToT strategies are inelastic across compute levels.
-
Scalable Token-Level Hallucination Detection in Large Language Models
TokenHD uses a scalable data synthesis engine and importance-weighted training to create token-level hallucination detectors that work on free-form text and scale from 0.6B to 8B parameters, outperforming larger reaso...
-
Shepherd: Enabling Programmable Meta-Agents via Reversible Agentic Execution Traces
Shepherd provides a reversible execution trace substrate for LLM agents that enables meta-agents to inspect and transform runs, yielding reported gains on coding and terminal benchmarks via supervision, counterfactual...
-
Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration
SPEX delivers 1.2-3x speedup on ToT algorithms via speculative path selection, dynamic budget allocation, and adaptive early termination, reaching up to 4.1x when combined with token-level speculative decoding.
-
Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration
SPEX accelerates Tree-of-Thought LLM reasoning 1.2-3x via speculative path selection, dynamic budget allocation across queries, and adaptive early termination, with up to 4.1x when combined with token speculative decoding.
-
APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation
APCD reduces LLM hallucinations by expanding decoding paths adaptively when entropy signals uncertainty and by contrasting divergent paths to control their interaction.
-
A$^2$TGPO: Agentic Turn-Group Policy Optimization with Adaptive Turn-level Clipping
A²TGPO improves RL policy optimization for multi-turn agentic LLMs by normalizing information gain within same-depth turn groups, rescaling cumulative advantages by sqrt of term count, and modulating clipping ranges p...
-
Online Self-Calibration Against Hallucination in Vision-Language Models
OSCAR exploits the generative-discriminative gap in LVLMs to build online preference data with MCTS and dual-granularity rewards for DPO-based calibration, claiming SOTA hallucination reduction and improved multimodal...
-
Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning
NPR trains LLMs to reason in parallel via self-distilled RL, delivering up to 24.5% performance gains and 4.6x speedups with 100% genuine parallel execution on reasoning benchmarks.
-
Less Diverse, Less Safe: The Indirect But Pervasive Risk of Test-Time Scaling in Large Language Models
Curtailing diversity in candidate pools for test-time scaling increases unsafe LLM outputs, as demonstrated by a reference-guided reduction protocol that evades standard safety classifiers across open and closed models.
-
Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts
Retrieval-of-Thought organizes prior reasoning into a thought graph for retrieval and reward-guided recombination, reducing output tokens by up to 40% and latency by 82% while preserving accuracy on reasoning benchmarks.
-
ChipSeek: Optimizing Verilog Generation via EDA-Integrated Reinforcement Learning
ChipSeek is a hierarchical-reward reinforcement learning framework with Curriculum-Guided Dynamic Policy Optimization that integrates EDA simulator feedback to improve LLM-generated RTL code on both functional correct...
-
ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
-
LMM-R1: Empowering 3B LMMs with Strong Reasoning Abilities Through Two-Stage Rule-Based RL
A two-stage RL framework first boosts text reasoning in 3B LMMs then adapts it to multimodal inputs, producing modest benchmark gains of 4.5-4.8%.
-
Efficient Multi-Agent System Training with Data Influence-Oriented Tree Search
DITS replaces Q-value guidance in MCTS with influence scores for synthetic data synthesis in multi-agent LLM training, claiming better efficiency and performance on eight datasets.
-
Inference-Time Scaling for Diffusion Models beyond Scaling Denoising Steps
Diffusion models improve generation quality via inference-time search over noise candidates guided by verifiers and algorithms, yielding gains beyond denoising step scaling on class- and text-conditioned benchmarks.
-
Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning
Survey mapping RL techniques onto LLM training and highlighting gaps in value-based, off-policy, and bootstrapping methods.
-
Test-Time Scaling in Multimodal Foundation Models: A Comprehensive Survey of Generation and Reasoning
A survey of test-time scaling for multimodal foundation models that introduces a three-way taxonomy of sampling, feedback, and search approaches along with applications and benchmarks.
-
Shepherd: Enabling Programmable Meta-Agents via Reversible Agentic Execution Traces
Shepherd is a runtime system that formalizes meta-agent operations via typed execution traces, enabling fast forking and demonstrated improvements in agent intervention, optimization, and training on benchmarks.
-
APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation
APCD adaptively branches LLM decoding paths based on token entropy and contrasts divergent paths to improve factual accuracy while preserving efficiency.
-
Not All Rollouts are Useful: Down-Sampling Rollouts in LLM Reinforcement Learning
PODS applies max-variance down-sampling to GRPO rollouts in LLM RLVR, delivering at least 1.7x faster training to peak test accuracy on reasoning benchmarks.
-
ScaleAcross Explorer: Exploring Communication Optimization for Scale-Across AI Model Training
ScaleAcross Explorer jointly optimizes three design dimensions for scale-across training and reports up to 64.62% speedups over production baselines and 37.59% over prior art in testbed and simulation experiments.
-
A Survey of Scaling in Large Language Model Reasoning
A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.
-
From System 1 to System 2: A Survey of Reasoning Large Language Models
The survey organizes the shift of LLMs toward deliberate System 2 reasoning, covering model construction techniques, performance on math and coding benchmarks, and future research directions.
-
Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models
The paper surveys reinforced reasoning techniques for LLMs, covering automated data construction, learning-to-reason methods, and test-time scaling as steps toward Large Reasoning Models.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.