BottleCap AI Releases ThinkingCap-Qwen3.8-27B: 37.2% Fewer Thinking Tokens at a 0.86pp Accuracy Cost
4 min
Contrastive-LM Releases CLM-8B: An Open System One Model That Scores Agent Actions Up to 9× Faster Than Jev
2 sources 4 min
Cross-Scale Transfer Learning for Depression Severity Prediction: From PHQ-8 to HAMD-17 Across Languages and Clinical Paradigms
When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
When EOS Tokens Disagree: Understanding Length Inflation in On-Policy Distillation
What Does Privileged Information Add to On-Policy Self-Distillation?
1
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
The Router Within: Eliciting Native Skill Routing from a Frozen LLM
1
Show HN: Nari Qwen3-TTS and Qwen3-ASR – High accuracy, low latency and cost
3 min1
Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models
Kraken: LLM-based Speech-to-Speech Translation via Low-bitrate VQ and Dual-path Source Conditioning
1
Everything in Moderation: Per-Domain Coverage Optima and Alignment-Resistant Domain Gaps in Multi-Domain Mid-Training
1
ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation
1
ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation
Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks
1
Signed Rescue Routing: Harm-Aware Cascades for Efficient LLM Inference
1
FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience
Graph Machine: Towards Better Pretraining via Edges
Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye
15 min1