CoolFace
Modelpublic

imdatta0/qwen3-4b-swegym-moto-kl02-multitrain-grpo-tfcov10-b04-lr1e6-s12-adapter

sourceHugging Faceupdated 4mo agoView on Hugging Face
0likes5downloads
Model Card

Qwen3-4B SWE-Gym moto KL02 target-file coverage ablation adapter

This repository contains a PEFT LoRA adapter checkpoint from the Qwen3 SWE-Gym moto investigation.

Negative ablation: target-file coverage bonus did not improve held-out greedy over the inherited KL02 adapter.

Local source checkpoint before upload:

/mnt/disks/unslothai/datta0/cache/qwen3-grpo-patch/20260604233602swegymq4b-kl02-multitrain-grpo-tfcov10-b04-lr1e6-s12b90e3a9/checkpoints/best_holdout

Use with the base model unsloth/Qwen3-4B-Instruct-2507.