PILOT in the Loop: Live Self-Improvement for Long-Horizon Agents Paper • 2608.26530 • Published 2 days ago • 25
SWE-Prune-Moe Collection Qwen3-Coder-Next MoE models pruned for software engineering using REAP, gate-based, and activation-based expert importance. • 3 items • Updated 15 days ago
RL4TG Collection Qwen2.5-3B checkpoints for Defects4J unit-test generation with Online Policy Distillation and GRPO. • 17 items • Updated Jul 25