BottleCap AI Releases ThinkingCap-Qwen3.8-27B: 37.2% Fewer Thinking Tokens at a 0.86pp Accuracy Cost
4 min
Train Where the Quantized Model Goes: On-Policy Distillation for Low-Bit Reasoning
Beyond Repeated Sampling: Learning Search Policies for LLM Reasoning
Capable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models
Reasoning Topology Matters: A Controlled Study of LLM-Based Cybersecurity Analysis
1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation
Calibrating Teacher--Student Discrepancy for On-Policy Distillation
What Does Privileged Information Add to On-Policy Self-Distillation?
1
When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models
1
Discrete Beckmann Transport Models for One-Step Language Modeling and Reasoning
1
Learning to Coach for Experiential Learning
2
Register Tokens for Bounded-State Reasoning in Diffusion Language Models
1
Omni-Streaming Thinking
1
Training Specialist Models without Reasoning Trajectories for Domain Expert Distillation
StepAudio 3 Realtime Technical Report
CanvasAnneal: Curriculum Reinforcement Learning for Diffusion Language Models
1
RetroThinker: Enabling Retrospective Thinking in Speech LLMs
1
DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF — new model trending #30 on Hugging Face
15 min
Negative Self-Distillation: Learning to Reason by Avoiding Flaws
1
Qwen 3.8 follows GPT-5.5 Pro reasoning prefills
Building Multilingual Bridges: Data Mixing as the Pillar of Generalization for In-Language Reasoning
1
Building Multilingual Bridges: Data Mixing as the Pillar of Generalization for In-Language Reasoning
1
On the Navier–Stokes Millennium Prize Problem
Difficulty-Adaptive Tree-Structured Policy Optimization for Expanding Reasoning Coverage in RLVR
2