Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability
Paper • 2610.08448 • Published • 153
taesiri PRO
taesiri
AI & ML interests
AGI ... one linear layer at a time
Recent Activity
upvoted a paper about 2 hours ago
Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability upvoted a paper about 2 hours ago
DuoMatching: Joint-Marginal Distribution Matching for Few-Step Video Generation upvoted a paper about 2 hours ago
EVISKILL: Grounding Skill Evolution in Replayable EvidenceOrganizations
upvoted 4
papers about 2 hours ago
upvoted a paper about 17 hours ago
upvoted 2
papers about 18 hours ago
upvoted 3
papers 1 day ago
OSWorld-Pro: Process-based Evaluation for Computer Use Agents
Paper • 2609.24890 • Published • 22
Dynamic Harness Search: Building Multi-Agent Systems Per-Query via Prediction
Paper • 2610.04137 • Published • 10
Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation
Paper • 2610.05608 • Published • 132
upvoted a paper 2 days ago
upvoted 6
papers 3 days ago
Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation
Paper • 2610.02788 • Published • 17
Octrees as an Explicit 3D Language
Paper • 2610.02388 • Published • 15
Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite
Paper • 2610.02826 • Published • 99
EditHero: A Benchmark for Long-Horizon Part-Level 3D Editing and Vibe Modeling
Paper • 2610.02298 • Published • 55
Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It
Paper • 2609.36585 • Published • 76
World Observer: Joint Actor-Observer Generation for Persistent World Modeling
Paper • 2610.02162 • Published • 85
upvoted 3
papers 6 days ago