POISE trains a lightweight probe on the actor's internal states to predict expected rewards for RLVR, matching DAPO performance on math benchmarks with lower compute by avoiding extra rollouts or critic models.
hub Canonical reference
Demystifying Long Chain-of-Thought Reasoning in LLMs
Canonical reference. 86% of citing Pith papers cite this work as background.
30
Pith papers citing it
Background
86% of classified citations
abstract
Scaling inference compute enhances reasoning in large language models (LLMs), with long chains-of-thought (CoTs) enabling strategies like backtracking and error correction. Reinforcement learning (RL) has emerged as a crucial method for developing these capabilities, yet the conditions under which long CoTs emerge remain unclear, and RL training requires careful design choices. In this study, we systematically investigate the mechanics of long CoT reasoning, identifying the key factors that enable models to generate long CoT trajectories. Through extensive supervised fine-tuning (SFT) and RL experiments, we present four main findings: (1) While SFT is not strictly necessary, it simplifies training and improves efficiency; (2) Reasoning capabilities tend to emerge with increased training compute, but their development is not guaranteed, making reward shaping crucial for stabilizing CoT length growth; (3) Scaling verifiable reward signals is critical for RL. We find that leveraging noisy, web-extracted solutions with filtering mechanisms shows strong potential, particularly for out-of-distribution (OOD) tasks such as STEM reasoning; and (4) Core abilities like error correction are inherently present in base models, but incentivizing these skills effectively for complex tasks via RL demands significant compute, and measuring their emergence requires a nuanced approach. These insights provide practical guidance for optimizing training strategies to enhance long CoT reasoning in LLMs. Our code is available at: https://github.com/eddycmu/demystify-long-cot.
hub tools
citation-role summary
background 6
dataset 1
citation-polarity summary
representative citing papers
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
Chain-of-Zoom factorizes extreme super-resolution into an autoregressive sequence of intermediate scales using a reused backbone model plus GRPO-tuned multi-scale VLM prompts.
SWE-RL uses RL on software evolution data to train LLMs achieving 41% on SWE-bench Verified with generalization to other reasoning tasks.
CLORE augments correct on-policy rollouts by deleting repetitive and irrelevant segments then optimizes with auxiliary DPO to improve accuracy-efficiency trade-off on math benchmarks.
SCOLAR fixes information gain collapse in latent visual reasoning by generating independent auxiliary visual tokens via a detransformer, extending acceptable CoT length over 30x and delivering +14.12% gains on reasoning benchmarks.
BET reduces reasoning tokens by about 55% on average while improving performance across benchmarks by learning to short-solve easy queries, fold early on unsolvable ones, and preserve budget for hard solvable queries.
VLA-ATTC equips VLA models with adaptive test-time compute via an uncertainty clutch and relative action critic, cutting failure rates by over 50% on LIBERO-LONG.
PERSA combines RLHF with selective parameter-efficient updates to top transformer layers, raising style alignment scores from 35% to 96% on code feedback benchmarks while holding correctness near 100%.
rePIRL learns effective process reward models for LLM reasoning via a dual policy-PRM update process inspired by inverse RL, unifying online and offline methods with reported gains over prior approaches on math and coding datasets.
Group Causal Counterfactual Policy Optimization trains LLMs on generalizable reasoning by defining episodic rewards for counterfactual robustness and transferability then optimizing the policy with token-level advantages.
Geo-R1 uses indirect proxy rewards from cross-view alignment with geolocation metadata to drive reinforcement learning, enabling zero-shot geospatial reasoning that transfers across 25+ tasks and sometimes exceeds supervised specialists.
CoT reasoning is a brittle mirage governed by distribution discrepancy between training and test data, demonstrated via controlled experiments in the new DataAlchemy environment.
LongWriter-Zero applies RL from a base model with specialized rewards for length, quality, and structure to outperform SFT baselines and larger models on long-writing benchmarks.
ViGoRL introduces visually grounded RL that anchors reasoning steps to image coordinates and uses multi-turn zooming to outperform standard RL and supervised baselines on spatial and GUI reasoning benchmarks.
A simple PPO-based RL training pipeline on base models scales reasoning performance and response length, outperforming prior work on math and science benchmarks with one-tenth the training steps.
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
HAB applies coarse-to-fine budgeting to LLM reasoning, predicting per-problem depth and learning intra-step token budgets via PPL comparisons and adaptive Pareto optimization, yielding higher accuracy and lower token use than standard CoT on GSM8K and MATH500.
VisionPulse is a step-wise visual token pruning method for LMMs that retains 5% of tokens per step, shortens reasoning traces by 11.2%, and maintains accuracy.
SLAT applies segment-level adaptive trimming in RL to reduce CoT reasoning length by 50% while maintaining competitive accuracy on benchmarks.
StraTA improves LLM agent success rates to 93.1% on ALFWorld and 84.2% on WebShop by sampling a compact initial strategy and training it jointly with action execution via hierarchical GRPO-style rollouts.
Video generation models demonstrate competitive multimodal reasoning on a new benchmark, matching or exceeding VLMs on visual puzzles and achieving 92% on MATH and 69.2% on MMMU.
Self-aligned reward uses relative perplexity differences to encourage concise, query-specific reasoning in LLMs, yielding 4% accuracy gains and 30% lower inference cost when added to PPO or GRPO.
ThinkAct introduces reinforced visual latent planning in a dual VLA system to enable better long-horizon reasoning and adaptation for embodied tasks.
citing papers explorer
-
Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States
POISE trains a lightweight probe on the actor's internal states to predict expected rewards for RLVR, matching DAPO performance on math benchmarks with lower compute by avoiding extra rollouts or critic models.
-
Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
-
Chain-of-Zoom: Extreme Super-Resolution via Scale Autoregression and Preference Alignment
Chain-of-Zoom factorizes extreme super-resolution into an autoregressive sequence of intermediate scales using a reused backbone model plus GRPO-tuned multi-scale VLM prompts.
-
SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution
SWE-RL uses RL on software evolution data to train LLMs achieving 41% on SWE-bench Verified with generalization to other reasoning tasks.
-
CLORE: Content-Level Optimization for Reasoning Efficiency
CLORE augments correct on-policy rollouts by deleting repetitive and irrelevant segments then optimizes with auxiliary DPO to improve accuracy-efficiency trade-off on math benchmarks.
-
Self-Consistent Latent Reasoning: Long Latent Sequence Reasoning for Vision-Language Model
SCOLAR fixes information gain collapse in latent visual reasoning by generating independent auxiliary visual tokens via a detransformer, extending acceptable CoT length over 30x and delivering +14.12% gains on reasoning benchmarks.
-
Nice Fold or Hero Call: Learning Budget-Efficient Thinking for Adaptive Reasoning
BET reduces reasoning tokens by about 55% on average while improving performance across benchmarks by learning to short-solve easy queries, fold early on unsolvable ones, and preserve budget for hard solvable queries.
-
VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model
VLA-ATTC equips VLA models with adaptive test-time compute via an uncertainty clutch and relative action critic, cutting failure rates by over 50% on LIBERO-LONG.
-
PERSA: Reinforcement Learning for Professor-Style Personalized Feedback with LLMs
PERSA combines RLHF with selective parameter-efficient updates to top transformer layers, raising style alignment scores from 35% to 96% on code feedback benchmarks while holding correctness near 100%.
-
rePIRL: Learn PRM with Inverse RL for LLM Reasoning
rePIRL learns effective process reward models for LLM reasoning via a dual policy-PRM update process inspired by inverse RL, unifying online and offline methods with reported gains over prior approaches on math and coding datasets.
-
Towards Generalizable Reasoning: Group Causal Counterfactual Policy Optimization for LLM Reasoning
Group Causal Counterfactual Policy Optimization trains LLMs on generalizable reasoning by defining episodic rewards for counterfactual robustness and transferability then optimizing the policy with token-level advantages.
-
Unlocking Zero-Shot Geospatial Reasoning via Indirect Rewards
Geo-R1 uses indirect proxy rewards from cross-view alignment with geolocation metadata to drive reinforcement learning, enabling zero-shot geospatial reasoning that transfers across 25+ tasks and sometimes exceeds supervised specialists.
-
Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens
CoT reasoning is a brittle mirage governed by distribution discrepancy between training and test data, demonstrated via controlled experiments in the new DataAlchemy environment.
-
LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning
LongWriter-Zero applies RL from a base model with specialized rewards for length, quality, and structure to outperform SFT baselines and larger models on long-writing benchmarks.
-
Grounded Reinforcement Learning for Visual Reasoning
ViGoRL introduces visually grounded RL that anchors reasoning steps to image coordinates and uses multi-turn zooming to outperform standard RL and supervised baselines on spatial and GUI reasoning benchmarks.
-
Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model
A simple PPO-based RL training pipeline on base models scales reasoning performance and response length, outperforming prior work on math and science benchmarks with one-tenth the training steps.
-
Characterize Then Distill: Mechanistic Reasoning in Large Output Spaces
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
-
Thinking Economically: A Hierarchical Framework for Adaptive-Complexity Reasoning in LLMs
HAB applies coarse-to-fine budgeting to LLM reasoning, predicting per-problem depth and learning intra-step token budgets via PPL comparisons and adaptive Pareto optimization, yielding higher accuracy and lower token use than standard CoT on GSM8K and MATH500.
-
VisionPulse: Dynamic Visual Sparsity for Efficient Multimodal Reasoning
VisionPulse is a step-wise visual token pruning method for LMMs that retains 5% of tokens per step, shortens reasoning traces by 11.2%, and maintains accuracy.
-
SLAT: Segment-Level Adaptive Trimming for Efficient CoT Reasoning
SLAT applies segment-level adaptive trimming in RL to reduce CoT reasoning length by 50% while maintaining competitive accuracy on benchmarks.
-
StraTA: Incentivizing Agentic Reinforcement Learning with Strategic Trajectory Abstraction
StraTA improves LLM agent success rates to 93.1% on ALFWorld and 84.2% on WebShop by sampling a compact initial strategy and training it jointly with action execution via hierarchical GRPO-style rollouts.
-
Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm
Video generation models demonstrate competitive multimodal reasoning on a new benchmark, matching or exceeding VLMs on visual puzzles and achieving 92% on MATH and 69.2% on MMMU.
-
Self-Aligned Reward: Towards Effective and Efficient Reasoners
Self-aligned reward uses relative perplexity differences to encourage concise, query-specific reasoning in LLMs, yielding 4% accuracy gains and 30% lower inference cost when added to PPO or GRPO.
-
ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning
ThinkAct introduces reinforced visual latent planning in a dual VLA system to enable better long-horizon reasoning and adaptation for embodied tasks.
-
Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models
A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.
-
The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes
A literature survey that introduces a taxonomy for LLM reasoning paradigms, analyzes methodological trends, and synthesizes failure modes from over 300 papers.
-
Phi-4-reasoning Technical Report
A 14B reasoning model trained via supervised fine-tuning on selected prompts and o3-mini traces, plus outcome RL, outperforms larger open models like DeepSeek-R1-Distill-Llama-70B on math, coding, planning and related benchmarks.
-
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.
-
Opportunities and Challenges of Large Language Models for Low-Resource Languages in Humanities Research
This survey paper identifies opportunities for LLMs in low-resource language humanities research along with challenges in data accessibility, model adaptability, and cultural sensitivity.
- Hint Tuning: Less Data Makes Better Reasoners