Add blog and final data

GRPO Phase 13: custom rollout_func for markdown JSON tool calls

SFT eval on 22-task held-out split — fill in leaderboard

Move SUPPORTS_CONCURRENT_SESSIONS from module-level to class attribute

Fix client._parse_result to unwrap {observation,reward,done} payload

Enable concurrent sessions in env for GRPO training

Attribute hand-curated Round-1 tasks to Finch + list ALL 119 tasks in openenv.yaml

Update openenv.yaml — full 119-task inventory + 32 enumerated entries

Phase 11.7: interactive Prev/Next/Play replay (was static wall of HTML)

Add raw_logs.txt + HF Job + adapter links to dashboard and README

Phase 11.6: Kimi-K2.5 best-run replays in dashboard

Fix blank Space iframe — base_path needs trailing slash for Gradio mount

Fix blank Gradio iframe — set root_path='/dashboard' on mount

Phase 11.5: Gradio dashboard at /dashboard (now the Space's base_path)

eval_lora: fix truncation drop-direction bug + add subprocess preflight

Phase 11: eval_lora.py — in-process SFT eval (no API, no WebSocket)

SFT run #2: 8K-context Qwen2.5-Coder-3B (qwen3b-office-sft-kimi-long)

Track PNGs via LFS so HF Space accepts the SFT plot

Phase 10.1: SFT log analyzer + Qwen2.5-Coder-3B training artifacts

Add SFT corpus + Kimi-K2.5 teacher run + Kimi eval run

train_sft: drop fp16, prefer bf16 (MPS-compatible without grad scaler)

Phase 10: SFT training script (Qwen2.5-Coder-3B + LoRA via TRL)

Phase 9.1: --skip-completed flag for cheap re-runs

Phase 9: hard early-submit gate at env layer (kills the exploit class)

Phase 8: SFT corpus builder with 6-filter pipeline

Parse Kimi K2/K2.5 native tool-call format in inference.py

Phase 7: close the 'submit source unchanged' exploit Kimi-K2.5 found

Round 2 README, Qwen2.5-Coder-3B baseline, missing data_pipeline pullers

Add graders

Add extended arena stuff

Update readme

Graduated code step rewards based on execution success and code substance

Enable web interface on HF Space

add finch attribution

Financial Task Environment — code execution with real xlsx

Initial commit