---
license: apache-2.0
base_model: Qwen/Qwen3-VL-8B-Instruct
library_name: peft
tags:
- lora
- process-reward-model
- multimodal
- chart-reasoning
- em-prm
---
# EM-PRM v2 — `a3_claims_s0`

LoRA adapter for **Qwen/Qwen3-VL-8B-Instruct** from the paper *EM-PRM: Evidence-Mediated Process Rewards for Robust Multimodal Reasoning* (EM-PRM v2 experiment ladder).

A3 claims-stage checkpoint (see planning/V2_PLAN.md in the dataset mirror for its pre-registration).

## Training
- LoRA rank 64, alpha 128, dropout 0.05, target modules down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj; vision tower frozen; bfloat16.
- Seed 0, learning rate 5e-05, effective batch 2×4, one epoch.
- Training data, pair sets and every gate artifact are in the mirror `RESEARCH-EMPRM/emprm-v2` (dataset repo; `results/runs_v2/train/a3_claims_s0/`) and the paper bundle under `backdata/`.

## Load
```python
from transformers import AutoModelForImageTextToText, AutoProcessor
from peft import PeftModel
base = AutoModelForImageTextToText.from_pretrained("Qwen/Qwen3-VL-8B-Instruct", dtype="bfloat16", device_map="cuda")
model = PeftModel.from_pretrained(base, "RESEARCH-EMPRM/emprm-v2-a3_claims_s0")
processor = AutoProcessor.from_pretrained("Qwen/Qwen3-VL-8B-Instruct")
```
`adapter_config.json` records the local path the adapter was trained from; pass the base model explicitly as above. Scoring prompts (bank extraction, claim extraction, claim support, ranking) are the ones in `work/scripts/eval_bon.py` of the mirror.

## Provenance
Trained  in the EM-PRM v2 repository; every number quoted in the paper is traceable to `planning/V2_PLAN.md` and the generated tables in the mirror.