AutoTTS discovers width-depth test-time scaling controllers through agentic search in a pre-collected trajectory environment, yielding better accuracy-cost tradeoffs than hand-designed baselines on math reasoning tasks at low cost.
hub Canonical reference
Improve Mathematical Reasoning in Language Models by Automated Process Supervision
Canonical reference. 94% of citing Pith papers cite this work as background.
63
Pith papers citing it
6
external citations · Pith
Background
94% of classified citations
abstract
Complex multi-step reasoning tasks, such as solving mathematical problems or generating code, remain a significant hurdle for even the most advanced large language models (LLMs). Verifying LLM outputs with an Outcome Reward Model (ORM) is a standard inference-time technique aimed at enhancing the reasoning performance of LLMs. However, this still proves insufficient for reasoning tasks with a lengthy or multi-hop reasoning chain, where the intermediate outcomes are neither properly rewarded nor penalized. Process supervision addresses this limitation by assigning intermediate rewards during the reasoning process. To date, the methods used to collect process supervision data have relied on either human annotation or per-step Monte Carlo estimation, both prohibitively expensive to scale, thus hindering the broad application of this technique. In response to this challenge, we propose a novel divide-and-conquer style Monte Carlo Tree Search (MCTS) algorithm named \textit{OmegaPRM} for the efficient collection of high-quality process supervision data. This algorithm swiftly identifies the first error in the Chain of Thought (CoT) with binary search and balances the positive and negative examples, thereby ensuring both efficiency and quality. As a result, we are able to collect over 1.5 million process supervision annotations to train Process Reward Models (PRMs). This fully automated process supervision alongside the weighted self-consistency algorithm is able to enhance LLMs' math reasoning performances. We improved the success rates of the instruction-tuned Gemini Pro model from 51\% to 69.4\% on MATH500 and from 86.4\% to 93.6\% on GSM8K. Similarly, we boosted the success rates of Gemma2 27B from 42.3\% to 58.2\% on MATH500 and from 74.0\% to 92.2\% on GSM8K. The entire process operates without any human intervention or supervision, making our method both financially and ...
hub tools
citation-role summary
background 16
method 1
citation-polarity summary
representative citing papers
PRISM is a contrastive, policy-aware training framework for process reward models that reduces false positives by 22% on PRMBench and boosts downstream accuracy up to 33% in Best-of-N selection by learning reliable relative comparisons instead of pointwise labels.
Prefix gain measured via student-model solve-rate improvement is used to train a Prefix Utility Model (PUM) that supplies stronger supervision than correctness-based process rewards for mathematical reasoning.
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 extraction per budget unit, showing empirical gains in optimization-like reasoning.
DistIL applies distributional DAgger with forward cross-entropy to achieve monotonic policy improvement and better Pass@N from rich feedback in RL for reasoning tasks.
ChemCoTBench-V2 is a new rule-verifiable benchmark with 5,620 samples across 18 tasks that evaluates LLM chemical reasoning traces using deterministic chemistry rules and reference traces rather than final answers alone.
Chunk-Level Guided Generation uses off-the-shelf large LLMs to score fixed-length chunks from small models via likelihoods, matching trained PRM performance on math benchmarks without reward-model training.
Unsupervised PRMs derived from LLM probabilities achieve up to 15% better error detection than LLM judges and match supervised PRMs in verification and RL tasks.
Presents verifiable counterfactual process supervision that generates annotated trajectories via template-aware error injection on symbolic chains, improving Best-of-8 reranking on logical reasoning benchmarks with preliminary math transfer.
DataPRM is an environment-aware generative process reward model that improves LLM data analysis agents by 7-11% on benchmarks via active verification and reflection-aware ternary rewards.
CoTEvol evolves CoT trajectories via reflective crossover and uncertainty-guided mutation to synthesize more accurate and diverse math reasoning data, outperforming distillation and search-based methods.
DuET uses dual execution of generated code and pseudocode with majority voting to achieve state-of-the-art test output prediction, boosting Pass@1 by 13.6 percentage points on LiveCodeBench.
GenAC introduces generative critics with chain-of-thought reasoning and in-context conditioning to improve value approximation and downstream RL performance in LLMs compared to value-based and value-free baselines.
RLSD mixes self-distillation for token-level policy difference magnitudes with RLVR for reliable update directions from response correctness to reach higher convergence and better training stability.
MTR-Bench is a new automated benchmark for multi-turn reasoning in LLMs covering diverse tasks and difficulty levels with 3600 instances.
o1-like models overthink easy tasks; self-training reduces compute use without accuracy loss on GSM8K, MATH500, GPQA, and AIME.
ORM-based test-time verification improves Text-to-SQL accuracy over heuristic selection by up to 4.33% on BIRD and 2.10% on Spider using automated labeling.
The log-probability ratio from RL post-training recovers the optimal advantage function, providing an effective free signal for test-time scaling, uncertainty estimation, and failure attribution in LLM agents.
Success Visitation Matching uses a discriminator to turn sparse outcome rewards into dense process rewards by matching visitations of successful episodes, provably preserving the optimal policy and speeding up robotic RL finetuning.
CATPO introduces an informativeness score F(T) and critique-guided healing for failed trees to improve efficiency and performance in tree-based RLVR, reaching 37.5% macro accuracy on math benchmarks.
TAPO corrects credit misassignment in RL for multimodal search agents by using tool parameter similarity to share advantages across equivalent actions.
ThoughtFold applies introspective redundancy detection within correct CoT trajectories to create sub-trajectory spectra, then uses masked preference optimization to penalize redundant explorations, yielding 56% token reduction on DeepSeek-R1-Distill-Qwen-7B while preserving accuracy.
LRMs show a large production-evaluation gap on the VAIR dataset with valid answers but invalid reasoning, driven by answer confirmation bias as evidenced by CoT analysis, linear probes, and causal patching.
Set-to-set distances on sentence embeddings provide a permutation-invariant reward signal that improves GRPO training and enables efficient test-time scaling for vision-language models generating chest X-ray reports.
citing papers explorer
-
LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling
AutoTTS discovers width-depth test-time scaling controllers through agentic search in a pre-collected trajectory environment, yielding better accuracy-cost tradeoffs than hand-designed baselines on math reasoning tasks at low cost.
-
The Hidden Bias of Process Reward Models:PRISM for Rewarding the Right Reasoning
PRISM is a contrastive, policy-aware training framework for process reward models that reduces false positives by 22% on PRMBench and boosts downstream accuracy up to 33% in Best-of-N selection by learning reliable relative comparisons instead of pointwise labels.
-
From Correctness to Utility: Gain-Based Prefix Evaluation for LLM Reasoning
Prefix gain measured via student-model solve-rate improvement is used to train a Prefix Utility Model (PUM) that supplies stronger supervision than correctness-based process rewards for mathematical reasoning.
-
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 extraction per budget unit, showing empirical gains in optimization-like reasoning.
-
Reinforcement Learning from Rich Feedback with Distributional DAgger
DistIL applies distributional DAgger with forward cross-entropy to achieve monotonic policy improvement and better Pass@N from rich feedback in RL for reasoning tasks.
-
From Answers to States: Verifiable Process-Level Evaluation of Chemical Reasoning in Large Language Models
ChemCoTBench-V2 is a new rule-verifiable benchmark with 5,620 samples across 18 tasks that evaluates LLM chemical reasoning traces using deterministic chemistry rules and reference traces rather than final answers alone.
-
Off-the-Shelf LLMs as Process Scorers: Training-Free Alternative to PRMs for Mathematical Reasoning
Chunk-Level Guided Generation uses off-the-shelf large LLMs to score fixed-length chunks from small models via likelihoods, matching trained PRM performance on math benchmarks without reward-model training.
-
Unsupervised Process Reward Models
Unsupervised PRMs derived from LLM probabilities achieve up to 15% better error detection than LLM judges and match supervised PRMs in verification and RL tasks.
-
Verifiable Counterfactual Supervision for Process Reward Models
Presents verifiable counterfactual process supervision that generates annotated trajectories via template-aware error injection on symbolic chains, improving Best-of-8 reranking on logical reasoning benchmarks with preliminary math transfer.
-
Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis
DataPRM is an environment-aware generative process reward model that improves LLM data analysis agents by 7-11% on benchmarks via active verification and reflection-aware ternary rewards.
-
CoTEvol: Self-Evolving Chain-of-Thoughts for Data Synthesis in Mathematical Reasoning
CoTEvol evolves CoT trajectories via reflective crossover and uncertainty-guided mutation to synthesize more accurate and diverse math reasoning data, outperforming distillation and search-based methods.
-
DuET: Dual Execution for Test Output Prediction with Generated Code and Pseudocode
DuET uses dual execution of generated code and pseudocode with majority voting to achieve state-of-the-art test output prediction, boosting Pass@1 by 13.6 percentage points on LiveCodeBench.
-
Bringing Value Models Back: Generative Critics for Value Modeling in LLM Reinforcement Learning
GenAC introduces generative critics with chain-of-thought reasoning and in-context conditioning to improve value approximation and downstream RL performance in LLMs compared to value-based and value-free baselines.
-
Self-Distilled RLVR
RLSD mixes self-distillation for token-level policy difference magnitudes with RLVR for reliable update directions from response correctness to reach higher convergence and better training stability.
-
MTR-Bench: A Comprehensive Benchmark for Multi-Turn Reasoning Evaluation
MTR-Bench is a new automated benchmark for multi-turn reasoning in LLMs covering diverse tasks and difficulty levels with 3600 instances.
-
Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs
o1-like models overthink easy tasks; self-training reduces compute use without accuracy loss on GSM8K, MATH500, GPQA, and AIME.
-
Test-Time Verification for Text-to-SQL via Outcome Reward Models
ORM-based test-time verification improves Text-to-SQL accuracy over heuristic selection by up to 4.33% on BIRD and 2.10% on Spider using automated labeling.
-
Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents
The log-probability ratio from RL post-training recovers the optimal advantage function, providing an effective free signal for test-time scaling, uncertainty estimation, and failure attribution in LLM agents.
-
Learning Process Rewards via Success Visitation Matching for Efficient RL
Success Visitation Matching uses a discriminator to turn sparse outcome rewards into dense process rewards by matching visitations of successful episodes, provably preserving the optimal policy and speeding up robotic RL finetuning.
-
CATPO: Critique-Augmented Tree Policy Optimization
CATPO introduces an informativeness score F(T) and critique-guided healing for failed trees to improve efficiency and performance in tree-based RLVR, reaching 37.5% macro accuracy on math benchmarks.
-
TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents
TAPO corrects credit misassignment in RL for multimodal search agents by using tool parameter similarity to share advantages across equivalent actions.
-
ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning
ThoughtFold applies introspective redundancy detection within correct CoT trajectories to create sub-trajectory spectra, then uses masked preference optimization to penalize redundant explorations, yielding 56% token reduction on DeepSeek-R1-Distill-Qwen-7B while preserving accuracy.
-
An Enigma of Artificial Reason: Investigating the Production-Evaluation Gap in Large Reasoning Models
LRMs show a large production-evaluation gap on the VAIR dataset with valid answers but invalid reasoning, driven by answer confirmation bias as evidenced by CoT analysis, linear probes, and causal patching.
-
SDR: Set-Distance Rewards for Radiology Report Generation
Set-to-set distances on sentence embeddings provide a permutation-invariant reward signal that improves GRPO training and enables efficient test-time scaling for vision-language models generating chest X-ray reports.
-
VeriGate: Verifier-Gated Step-Level Supervision for GRPO
VeriGate adds verifier-gated step-level supervision to GRPO via cumulated PRM rewards and group-normalized token advantages, raising accuracy 20% and 12% on 1.5B and 7B models on MATH and six benchmarks.
-
Credit Assignment with Resets in Language Model Reasoning
The paper introduces Random-Reset Policy Optimization (RRPO) and Self-Reset Policy Optimization (SRPO) that use resets to enable more precise credit assignment in RL for language model reasoning, with SRPO outperforming GRPO and RRPO across benchmarks.
-
Beyond the Frontier: Stochastic Backtracking for Efficient Test-Time Scaling
Stochastic backtracking with subpool selection and power backtrack SMC over persistent prefix pools improves accuracy per token in PRM-guided test-time scaling for mathematical reasoning.
-
What and When to Distill: Selective Hindsight Distillation for Multi-Turn Agents
SERL selectively reweights learning using task success and environment feedback to reach 90.0% success on ALFWorld and 80.1% on WebShop, outperforming RL and distillation baselines.
-
Step-wise Rubric Rewards for LLM Reasoning
SRaR attributes rubric items to specific steps via an LLM judge, normalizes per-step scores across rollouts, and combines them with outcome rewards via a decoupled advantage estimator, yielding 3.57-point accuracy gains on Qwen3-8B across math benchmarks.
-
Process Rewards with Learned Reliability
BetaPRM learns distributional step rewards with explicit reliability via Beta-Binomial modeling, enabling ACA that cuts token use by up to 33.57% while raising final-answer accuracy on reasoning benchmarks.
-
Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards
CIPO jointly optimizes standard RLVR rewards with correction samples derived from the model's own failed attempts, yielding better reasoning and self-correction on math and code benchmarks.
-
Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information
Anti-Self-Distillation reverses self-distillation signals via PMI to fix overconfidence on structural tokens, matching GRPO baseline accuracy 2-10x faster with up to 11.5 point gains across 4B-30B models.
-
Pause or Fabricate? Training Language Models for Grounded Reasoning
GRIL uses stage-specific RL rewards to train LLMs to detect missing premises, pause proactively, and resume grounded reasoning after clarification, yielding up to 45% better premise detection and 30% higher task success on insufficient math datasets.
-
Process Reward Models Meet Planning: Generating Precise and Scalable Datasets for Step-Level Rewards
PDDL planning problems are used to generate about one million precise reasoning steps for training Process Reward Models, and adding this data to existing datasets improves LLM performance on both mathematical and non-mathematical reasoning benchmarks.
-
ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models
Lightweight probes on LLM internal states verify reasoning steps as effectively as process reward models up to 810 times larger across math, planning, and QA tasks.
-
ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution
ShinkaEvolve improves sample efficiency in LLM-driven program evolution via parent sampling, code novelty rejection-sampling, and bandit LLM ensemble selection, achieving new SOTA circle packing with 150 samples and gains on math reasoning and competitive programming tasks.
-
Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
PROF curates RL training data via PRM-ORM consistency to improve both final-answer accuracy and intermediate reasoning quality while reducing reliance on strong process reward models.
-
InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
-
CoLD: Counterfactually-Guided Length Debiasing for Process Reward Models in Mathematical Reasoning
CoLD mitigates length bias in process reward models for mathematical reasoning via counterfactual guidance, length penalties, bias estimation, and joint training, improving step selection accuracy and conciseness on MATH500 and GSM-Plus while boosting downstream RL performance.
-
InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
InternVL3-78B sets a new open-source SOTA of 72.2 on MMMU via native joint multimodal pre-training, V2PE, MPO, and test-time scaling while remaining competitive with proprietary models.
-
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%.
-
The Lessons of Developing Process Reward Models in Mathematical Reasoning
Monte Carlo data synthesis for PRMs underperforms LLM-judge and human methods, Best-of-N evaluations suffer from process-outcome misalignment and score inflation, and consensus filtering yields better PRMs with higher data efficiency.
-
Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning
Process advantage verifiers trained to predict step-level progress under a distinct prover policy improve LLM reasoning accuracy by over 8% and sample efficiency by 5-6x over outcome reward models.
-
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.
-
Sakana Fugu Technical Report
Sakana Fugu trains LLM orchestrators using fine-tuning, evolutionary algorithms, and RL to build query-adaptive multi-agent scaffolds, claiming SOTA results on benchmarks including SWE-Bench Pro and GPQA-Diamond.
-
Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery
AI for mathematics is best described as a supervision ladder — final answers, programs, process rewards, proof-assistant kernels — culminating in verified-discovery workflows.
-
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.
-
Self-evolving LLM agents with in-distribution Optimization
Q-Evolve unifies automatic process-reward labeling via advantage estimation and behavior-proximal policy optimization inside an in-distribution RL loop to enable self-evolving LLM agents on interactive tasks.
-
Conformal Certification of Reasoning Trace Prefixes
CROP applies conformal prediction to certify the longest contiguous prefix of an LLM reasoning trace that is guaranteed to contain no annotated errors under an exchangeability assumption.
-
Prefix-Safe Bayesian Belief Tracking for LLM Reasoning Reliability:Separating Calibration from Ranking
SBBT separates Brier-score calibration gains from AUROC ranking gains in prefix-conditioned success estimation for LLM math reasoning, with structure-aware signals yielding up to +0.110 AUROC over baselines.