Jerry9999 commited on
Commit
fd1df4a
·
verified ·
1 Parent(s): 12bb32b

README: GRPO baseline

Browse files
Files changed (1) hide show
  1. README.md +4 -0
README.md CHANGED
@@ -11,3 +11,7 @@ Harbor Terminus-2 trials of Qwen3.5-9B checkpoints on OpenThoughts-TB-dev (70 ta
11
  - Score: lenient 53.36% / strict 51.83% (baseline GRPO 32 x 8 at the same step: 53.29% / 52.40%).
12
  - Each trial dir: `agent/trajectory.json` (full multi-turn conversation incl. thinking), `agent/trajectory.summarization-*` (context summaries), `agent/terminus_2.pane`, `verifier/` (test output, reward), `result.json`, `config.json`, `trial.log`.
13
  - API keys in configs are redacted. `AKIAIOSFODNN7EXAMPLE` in one task is AWS's documented example key, part of the task data.
 
 
 
 
 
11
  - Score: lenient 53.36% / strict 51.83% (baseline GRPO 32 x 8 at the same step: 53.29% / 52.40%).
12
  - Each trial dir: `agent/trajectory.json` (full multi-turn conversation incl. thinking), `agent/trajectory.summarization-*` (context summaries), `agent/terminus_2.pane`, `verifier/` (test output, reward), `result.json`, `config.json`, `trial.log`.
13
  - API keys in configs are redacted. `AKIAIOSFODNN7EXAMPLE` in one task is AWS's documented example key, part of the task data.
14
+
15
+ ## `tbdev/GRPO_default_32x8_iter67/`
16
+ - Model: Qwen3.5-9B after 68 async GRPO updates (32 x 8, staleness <= 2), the default setting, no teacher. Run `ag32x8_1102`, eval job 3767729. The baseline for K above (same step, same eval).
17
+ - Score: lenient 53.29% / strict 52.40%. Same file layout as above. Fake credentials such as `sk-secret-...` / `sekret-key-...` are task data.