STAR arm A2 (seed 42) — LoRA adapter

Authors: Chaitanya Medidar, Manasi Ghalsasi, Pouras Patil, Rushil Patil (Vivekanand Education Society's Institute of Technology, Mumbai)

Part of STAR: Execution-Grounded, Typed Recovery Supervision for Coding-Agent SFT. Code, paper materials and manifests: https://github.com/chaitanyamedidar/STAR

Arm A2: utility-only: the same median base-loss cut, no force-keep (ablation)
Base model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), not included; LoRA r=64, alpha=128, all attention+MLP projections.
Seed: 42. The training run manifest is run_manifest.json. Dataset (adapters, weights, labels, records): pourasrp/star_v1, DOI 10.57967/hf/11003. Per-token weight files are also in pourasrp/star-weights.

file	sha256
adapter_model.safetensors	`4dfe92988fc312d53e8196986e74fda7070f22eef30b8e55b77589079e782cf8`
adapter_config.json	`1448204e3648947def36dc5e38f69d4b55d124a95256cba00066a8a5c27d7414`

Verify with sha256sum adapter_model.safetensors.

from peft import PeftModel; from transformers import AutoModelForCausalLM
m = PeftModel.from_pretrained(AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B"), "pourasrp/star-A2")

License: Apache-2.0 (matches the GitHub repo). Training data came from SWE-Lego Real. No trajectory text is redistributed here.

Downloads last month: -

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for pourasrp/star-A2

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Adapter

(2274)

this model