IDPR is a response-conditioned inhibitory deliberation method that trains a controller on fast-slow outcome pairs to decide when to override LLM fast answers, improving accuracy from 47.90% to 48.92% with slow reasoning invoked on only 8.20% of a 5,000-example math test set.
hub
LIMO: Less is More for Reasoning
42 Pith papers cite this work. Polarity classification is still indexing.
42
Pith papers citing it
abstract
We challenge the prevailing assumption that complex reasoning in large language models (LLMs) necessitates massive training data. We demonstrate that sophisticated mathematical reasoning can emerge with only a few examples. Specifically, through simple supervised fine-tuning, our model, LIMO, achieves 63.3\% accuracy on AIME24 and 95.6\% on MATH500, surpassing previous fine-tuned models (6.5\% on AIME24, 59.2\% on MATH500) while using only 1\% of the training data required by prior approaches. Furthermore, LIMO exhibits strong out-of-distribution generalization, achieving a 45.8\% absolute improvement across diverse benchmarks, outperforming models trained on 100x more data. Synthesizing these findings, we propose the Less-Is-More Reasoning Hypothesis (LIMO Hypothesis): In foundation models where domain knowledge has been comprehensively encoded during pre-training, sophisticated reasoning can emerge through minimal but strategically designed demonstrations of cognitive processes. This hypothesis suggests that the threshold for eliciting complex reasoning is not dictated by task complexity but rather by two key factors: (1) the completeness of the model's pre-trained knowledge base and (2) the effectiveness of post-training examples in serving as "cognitive templates" that guide reasoning.
hub tools
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
Language models produce overcomplete reasoning traces where on average 46% of steps can be removed while preserving the answer in 86% of cases, with necessity concentrated in the top three steps.
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
GPS trains a small model on optimization history to predict prompt difficulty and select intermediate-difficulty diverse batches, yielding better training efficiency, final performance, and test-time allocation than baselines on reasoning benchmarks.
Parallel thinking in LLMs suffers from overscaling where fixed global budgets waste samples; LanBo predicts per-sample budgets from latent states to raise utilization without hurting accuracy.
A single LLM improves its own reasoning by self-distilling from privileged verified traces as teacher to its question-only student policy, outperforming off-policy distillation and RL on math benchmarks with better token efficiency.
LLMs show strong user bias in role-tagged contexts that is amplified by preference alignment and can be reduced or controlled through targeted fine-tuning and DPO.
SEAR trains one LLM via adversarial process rewards to explore harmful reasoning paths but flip to safe outputs, reducing over-refusal while preserving safety.
The LLM-as-Environment-Engineer framework lets the policy model redesign its own RL environments on the new MAPF-FrozenLake testbed, outperforming larger models and fixed baselines with Qwen3-4B.
CoT SFT disrupts long-range routing in hybrid models via changes to W_Q and W_K; QK-Restore restores pre-SFT projections to recover NIAH performance.
LaSR improves context-aware terminology recognition in speech LLMs by aligning latent CoT supervision on acoustic regions and introducing latent reasoning periods, shown on a new academic corpus to outperform standard fine-tuning without added latency.
PPC adds a preplan stage to the question-plan-CoT paradigm, achieving best results on 39 of 40 metrics across five math benchmarks with no added inference tokens.
DARE co-evolves difficulty estimation and policy in RL for LLMs to improve training efficiency, final performance, and inference speed by using tailored strategies for different difficulty levels.
SIAM achieves state-of-the-art whole-head MRI segmentation of 16 structures including extra-cerebral tissues by training on synthetic data from just six manual templates, matching or exceeding prior methods on 301 scans across eight heterogeneous datasets.
A large model generates a compact reasoning signal that a small model uses to solve tasks, reducing the large model's output tokens by up to 60% on benchmarks like AIME and GPQA.
HEAL mitigates entropy collapse in few-shot RLVR by selectively adding general-domain data and aligning trajectory-level entropy dynamics, matching full-shot performance with 32 target samples.
Recovering an orthogonal basis from model activations yields a model-native skill characterization that improves reasoning Pass@1 by up to 41% via targeted data selection and supports inference steering, outperforming human-characterized alternatives.
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.
CURE-MED pairs a new 13-language medical reasoning benchmark with curriculum RL to raise logical correctness to 70% and language consistency to 95% at 32B scale while outperforming baselines.
Selecting a short informative reference segment using audio diversity, lip amplitude, and viewpoint criteria achieves comparable personalized 3D talking face quality while reducing processing and training time by over 5x.
A Dirichlet-prior Bayesian estimator for model success probability replaces Pass@k, delivering faster-converging and more stable rankings with credible intervals on math benchmarks.
Fin-PRM is a domain-specialized process reward model that supplies binary step-level and trajectory-level supervision signals for financial reasoning in LLMs and outperforms general PRMs on CFLUE and FinQA benchmarks.
WebSailor trains open-source web agents to match proprietary performance on complex information-seeking tasks by generating high-uncertainty scenarios and using a new RL method called DUPO.
LUFFY mixes off-policy reasoning traces into RLVR training via Mixed-Policy GRPO and regularized importance sampling, delivering over 6-point gains on math benchmarks and enabling training of weak models where on-policy RLVR fails.
citing papers explorer
-
When to Think Deeply: Inhibitory Deliberation for LLM Reasoning
IDPR is a response-conditioned inhibitory deliberation method that trains a controller on fast-slow outcome pairs to decide when to override LLM fast answers, improving accuracy from 47.90% to 48.92% with slow reasoning invoked on only 8.20% of a 5,000-example math test set.
-
Uncovering the Representation Geometry of Minimal Cores in Overcomplete Reasoning Traces
Language models produce overcomplete reasoning traces where on average 46% of steps can be removed while preserving the answer in 86% of cases, with necessity concentrated in the top three steps.
-
Logic-Regularized Verifier Elicits Reasoning from LLMs
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
-
Small Generalizable Prompt Predictive Models Can Steer Efficient RL Post-Training of Large Reasoning Models
GPS trains a small model on optimization history to predict prompt difficulty and select intermediate-difficulty diverse batches, yielding better training efficiency, final performance, and test-time allocation than baselines on reasoning benchmarks.
-
On the Overscaling Curse of Parallel Thinking: System Efficacy Contradicts Sample Efficiency
Parallel thinking in LLMs suffers from overscaling where fixed global budgets waste samples; LanBo predicts per-sample budgets from latent states to raise utilization without hurting accuracy.
-
Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models
A single LLM improves its own reasoning by self-distilling from privileged verified traces as teacher to its question-only student policy, outperforming off-policy distillation and RL on math benchmarks with better token efficiency.
-
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.
-
From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning
The LLM-as-Environment-Engineer framework lets the policy model redesign its own RL environments on the new MAPF-FrozenLake testbed, outperforming larger models and fixed baselines with Qwen3-4B.
-
Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It
CoT SFT disrupts long-range routing in hybrid models via changes to W_Q and W_K; QK-Restore restores pre-SFT projections to recover NIAH performance.
-
LaSR: Context-Aware Speech Recognition via Latent Reasoning
LaSR improves context-aware terminology recognition in speech LLMs by aligning latent CoT supervision on acoustic regions and introducing latent reasoning periods, shown on a new academic corpus to outperform standard fine-tuning without added latency.
-
Knowing What to Solve Before How: Preplan Empowered LLM Mathematical Reasoning
PPC adds a preplan stage to the question-plan-CoT paradigm, achieving best results on 39 of 40 metrics across five math benchmarks with no added inference tokens.
-
DARE: Difficulty-Adaptive Reinforcement Learning with Co-Evolved Difficulty Estimation
DARE co-evolves difficulty estimation and policy in RL for LLMs to improve training efficiency, final performance, and inference speed by using tailored strategies for different difficulty levels.
-
SIAM: Head and Brain MRI Segmentation from Few High-Quality Templates via Synthetic Training
SIAM achieves state-of-the-art whole-head MRI segmentation of 16 structures including extra-cerebral tissues by training on synthetic data from just six manual templates, matching or exceeding prior methods on 301 scans across eight heterogeneous datasets.
-
When Less is Enough: Efficient Inference via Collaborative Reasoning
A large model generates a compact reasoning signal that a small model uses to solve tasks, reducing the large model's output tokens by up to 60% on benchmarks like AIME and GPQA.
-
HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment
HEAL mitigates entropy collapse in few-shot RLVR by selectively adding general-domain data and aligning trajectory-level entropy dynamics, matching full-shot performance with 32 target samples.
-
Characterizing Model-Native Skills
Recovering an orthogonal basis from model activations yields a model-native skill characterization that improves reasoning Pass@1 by up to 41% via targeted data selection and supports inference steering, outperforming human-characterized alternatives.
-
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.
-
CURE-Med: Curriculum-Informed Reinforcement Learning for Multilingual Medical Reasoning
CURE-MED pairs a new 13-language medical reasoning benchmark with curriculum RL to raise logical correctness to 70% and language consistency to 95% at 32B scale while outperforming baselines.
-
Trust Region On-Policy Distillation
TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.
-
Sample-Efficient Post-Training for LEGO Spatial-Physics Reasoning
PVPO is a sample-efficient RL method that improves semantic, geometric, and physical quality in LLM LEGO assembly generation by mitigating the PhysHack failure mode where validity alone fails to ensure fidelity.
-
TPMM-DPO: Trajectory-aware Preference-guided Model Merging for Iterative Direct Preference Optimization
TPMM-DPO applies trajectory-aware learned-weight merging of prior policy models to stabilize iterative DPO against preference noise accumulation.
-
Data Difficulty and the Generalization--Extrapolation Tradeoff in LLM Fine-Tuning
Optimal data difficulty for LLM supervised fine-tuning shifts toward harder examples as data budget increases due to the generalization-extrapolation tradeoff.
-
Learning to Foresee: Unveiling the Unlocking Efficiency of On-Policy Distillation
On-policy distillation gains efficiency from early foresight in module allocation and update directions, which the proposed EffOPD method exploits for 3x faster training with comparable performance.
- SUPERNOVA: Eliciting General Reasoning in LLMs with Reinforcement Learning on Natural Instructions
- Which Reasoning Trajectories Teach Students to Reason Better? A Simple Metric of Informative Alignment