Combined views
4.5K
2 posts, first seen 5h ago
36 likes
4.5K
2 posts, first seen 5h ago
Users praise the TRACE method as impressive because its self-diagnosis capability lets AI agents understand failures rather than just optimizing for success.
Based on 2 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@Azaliamirh the self-diagnosis part is the real breakthrough. Most approaches optimize for success, not for understanding why they failed. That meta-awareness loop is what separates toy agents from real ones.
@Azaliamirh impressive; looks like OSS is doing well
Check out TRACE, a new self-improvement approach where the agent identifies the missing capabilities behind its own failures and trains itself to address them. TRACE-trained Qwen3.6-27B reaches 73.2% on SWE-bench Verified, outperforming much larger models like Codex 5.2 and GLM 5, while beating GRPO and GEPA with <1/4 the training rollouts. By contrasting successful and failed trajectories, TRACE identifies its own weaknesses (such as bug localization or retrieval of the correct doc) and creates new synthetic environments to fix them. The result is a transferable and sample-efficient synthetic env / data generation + fine-tuning pipeline for agentic tasks. Great work led by @TarunSures41845 and @hangoo_kang!
A TRACE-trained Qwen 27B model scored 73.2% on SWE-bench Verified.
Users praise the TRACE method as impressive because its self-diagnosis capability lets AI agents understand failures rather than just optimizing for success.
Based on 2 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.