File size: 1,531 Bytes
82cf7e4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 | ---
license: apache-2.0
base_model: Qwen/Qwen3.6-27B
tags: [debugging, code, reinforcement-learning, grpo, precisecoder]
---
# PreciseCoder-27B
Qwen3.6-27B post-trained with reinforcement learning for **precise debugging**: given buggy Python
code, produce a minimal, correct fix (PreciseCoder project). Best-validation checkpoint, global step 328 (2 epochs).
**Recipe.** GRPO from the base model (no SFT warm-start) with reward = (unit + 0.5路F1) / 1.5, where F1 is the
edit-block precision/recall F1 against the ground-truth diff (PDB evaluator), on the PDB train split (buggy inputs only).
Trained with verl on 2 nodes (n=6 rollouts, temperature 1.0, lr 1e-6, batch 32, max response 6144 tokens,
KL loss 0.005, rollout TP=2). Minimal-edit prompt template.
**Eval** (PDB benchmark, BigCodeBench + LiveCodeBench debugging tasks):
| eval | result |
|---|---|
| pdb_test (buggy inputs, 4k think + 2k answer) | unit 0.777 路 precision 0.784 路 recall 0.850 路 edit lines 2.7 |
| gt_holdout (1512 clean inputs, 16k think + 4k answer) | no-op 47.2% 路 damage 10.2% 路 unit pass 89.7% |
| PDB-Wild (228 SWE-smith repo-level bugs) | unit 0.531 路 precision 0.544 路 recall 0.622 路 F1 0.559 路 edit lines 9.4 |
`no-op` = share of clean (bug-free) inputs returned unchanged; `damage` = share broken by an edit.
Computed over answers with an extractable code block.
Load with `transformers` / vLLM (>=0.17) as a standard Qwen3.6 model (thinking enabled by default).
Source: verl experiment `qwen36-27b-GRPO_0.5F1` (Delta, Sep 2026).
|