Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning
Paper • 2510.03259 • Published • 57
LLM
-
-
Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense
Paper • 2510.07242 • Published • 30 -
First Try Matters: Revisiting the Role of Reflection in Reasoning Models
Paper • 2510.08308 • Published • 24 -
Low-probability Tokens Sustain Exploration in Reinforcement Learning with Verifiable Reward
Paper • 2510.03222 • Published • 76 -
Latent Refinement Decoding: Enhancing Diffusion-Based Language Models by Refining Belief States
Paper • 2510.11052 • Published • 53 -
RLFR: Extending Reinforcement Learning for LLMs with Flow Environment
Paper • 2510.10201 • Published • 37 -
Making Mathematical Reasoning Adaptive
Paper • 2510.04617 • Published • 23 -
Demystifying Reinforcement Learning in Agentic Reasoning
Paper • 2510.11701 • Published • 34 -
Are Large Reasoning Models Interruptible?
Paper • 2510.11713 • Published • 5 -
QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMs
Paper • 2510.11696 • Published • 183 -
Deep Self-Evolving Reasoning
Paper • 2510.17498 • Published • 12 -
Continuous Autoregressive Language Models
Paper • 2510.27688 • Published • 74 -
Higher-order Linear Attention
Paper • 2510.27258 • Published • 15 -
Limits of Generalization in RLVR: Two Case Studies in Mathematical Reasoning
Paper • 2510.27044 • Published • 6 -
Why Language Models Hallucinate
Paper • 2509.04664 • Published • 200 -
Reverse-Engineered Reasoning for Open-Ended Generation
Paper • 2509.06160 • Published • 151 -
MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing
Paper • 2509.22186 • Published • 177 -
FlowRL: Matching Reward Distributions for LLM Reasoning
Paper • 2509.15207 • Published • 119 -
Towards a Unified View of Large Language Model Post-Training
Paper • 2509.04419 • Published • 77 -
Variational Reasoning for Language Models
Paper • 2509.22637 • Published • 70 -
Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models
Paper • 2509.06949 • Published • 57 -
Beyond the Exploration-Exploitation Trade-off: A Hidden State Approach for LLM Reasoning in RLVR
Paper • 2509.23808 • Published • 47 -
Sequential Diffusion Language Models
Paper • 2509.24007 • Published • 47 -
Every Token Counts: Generalizing 16M Ultra-Long Context in Large Language Models
Paper • 2511.23319 • Published • 25 -
GLM-5: from Vibe Coding to Agentic Engineering
Paper • 2602.15763 • Published • 205 -
Experiential Reinforcement Learning
Paper • 2602.13949 • Published • 76 -
Baichuan-M3: Modeling Clinical Inquiry for Reliable Medical Decision-Making
Paper • 2602.06570 • Published • 61 -
Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability
Paper • 2604.06628 • Published • 330 -
Self-Distilled RLVR
Paper • 2604.03128 • Published • 181 -
The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook
Paper • 2604.02029 • Published • 152 -
TriAttention: Efficient Long Reasoning with Trigonometric KV Compression
Paper • 2604.04921 • Published • 117 -
Can LLMs Learn to Reason Robustly under Noisy Supervision?
Paper • 2604.03993 • Published • 42 -
Large Language Models Explore by Latent Distilling
Paper • 2604.24927 • Published • 74 -
Why Fine-Tuning Encourages Hallucinations and How to Fix It
Paper • 2604.15574 • Published • 26 -
Hallucinations Undermine Trust; Metacognition is a Way Forward
Paper • 2605.01428 • Published • 24 -
Co-Evolving Policy Distillation
Paper • 2604.27083 • Published • 68 -
Efficient Training on Multiple Consumer GPUs with RoundPipe
Paper • 2604.27085 • Published • 47 -
Learning from Noisy Preferences: A Semi-Supervised Learning Approach to Direct Preference Optimization
Paper • 2604.24952 • Published • 6 -
Can RL Teach Long-Horizon Reasoning to LLMs? Expressiveness Is Key
Paper • 2605.06638 • Published • 16 -
Continuous Latent Diffusion Language Model
Paper • 2605.06548 • Published • 87 -
RouteProfile: Elucidating the Design Space of LLM Profiles for Routing
Paper • 2605.00180 • Published • 30 -
Long Context Pre-Training with Lighthouse Attention
Paper • 2605.06554 • Published • 31 -
Boosting Reinforcement Learning with Verifiable Rewards via Randomly Selected Few-Shot Guidance
Paper • 2605.15012 • Published • 4 -
BEAM: Binary Expert Activation Masking for Dynamic Routing in MoE
Paper • 2605.14438 • Published • 5 -
Process Rewards with Learned Reliability
Paper • 2605.15529 • Published • 53 -
It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs
Paper • 2605.20258 • Published • 30 -
The Unlearnability Phenomenon in RLVR for Language Models
Paper • 2605.16787 • Published • 7 -
ThriftAttention: Selective Mixed Precision for Long-Context FP4 Attention
Paper • 2605.23081 • Published • 41 -
Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth
Paper • 2605.25052 • Published • 14 -
Language Models Need Sleep
Paper • 2605.26099 • Published • 14 -
Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language Models
Paper • 2605.25189 • Published • 4 -
How Far Will They Go? Red-Teaming Online Influence with Large Language Models
Paper • 2605.22880 • Published • 6 -
Decoding the Critique Mechanism in Large Reasoning Models
Paper • 2603.16331 • Published -
DVAO: Dynamic Variance-adaptive Advantage Optimization for Multi-reward Reinforcement Learning
Paper • 2605.25604 • Published • 139 -
Self-Improving Language Models with Bidirectional Evolutionary Search
Paper • 2605.28814 • Published • 62 -
Triplet-Block Diffusion RWKV
Paper • 2605.25969 • Published • 27 -
Less is More: Early Stopping Rollout for On-Policy Distillation
Paper • 2605.27028 • Published • 15 -
Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention
Paper • 2605.29548 • Published • 13 -
On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
Paper • 2606.02437 • Published • 241 -
Rethinking the Divergence Regularization in LLM RL
Paper • 2606.09821 • Published • 34 -
Do Thinking Tokens Help with Safety?
Paper • 2606.25013 • Published • 1 -
Improved Large Language Diffusion Models
Paper • 2606.25331 • Published • 43 -
AsyncOPD: How Stale Can On-Policy Distillation Be?
Paper • 2606.24143 • Published • 31 -
From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement
Paper • 2607.23802 • Published • 93 -
Weak-to-Strong On-Policy Distillation
Paper • 2607.26246 • Published • 50 -
Enhancing Rubric-based RL via Self-Distillation
Paper • 2607.18082 • Published • 22 -
Constitutional Midtraining: Content Presence Drives Alignment Gains
Paper • 2607.26654 • Published • 6 -
DiffusionGemma Technical Report
Paper • 2608.00146 • Published • 29