LoRA adapter for Qwen2.5-VL-7B-Instruct, trained with GRPO with a temporal-IoU reward at 2 fps plus the placebo-controlled term (λ = 0.01), seed 0.
Released with the paper Accuracy Is Not Evidence: Measuring and Training Evidence Dependence in Video Temporal Grounding (ICLR 2027 submission, under review). Code, data identifiers and evaluation scripts are in iclropen/evidence-audit-code.
Trained for the revised version of the paper, where it appears in Tables 23, 26–29, 32, 33, 37 and 38 (Appendix F).
Load it with transformers 4.51 (tested 4.51.3, peft 0.20).
Adapters saved under transformers 4.51 and 5 use different module names. Loading one under the other version raises no error but leaves the LoRA weights unused, so use the version given here. tools/check_lora_keys.py in the code repository checks that an adapter actually loaded.
The transformers 4.51 runs used gradient checkpointing, under which the LoRA weights of the vision tower received no gradient. They are zero, so these adapters change the language model only.
# pip install "transformers==4.51.3" "peft==0.20.0" qwen-vl-utils
import torch
from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
from peft import PeftModel
base = "Qwen/Qwen2.5-VL-7B-Instruct"
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(base, torch_dtype=torch.bfloat16, device_map="auto")
processor = AutoProcessor.from_pretrained(base)
model = PeftModel.from_pretrained(model, "iclropen/evidence-audit-qwen7b-iougrpo-pc-lam001-s0")
The evaluation and audit scripts of the code repository take --model <base> --adapter <this repository or folder>.
GRPO with a temporal-IoU reward at 2 fps, with the settings of iclropen/evidence-audit-qwen7b-iougrpo-2fps-s0, plus the placebo-controlled term of the paper (Section 4) with weight λ.
| Setting | Value |
|---|---|
| λ | 0.01 |
| KL coefficient | 0.04 |
| Samples per question | 8 |
| Questions per step | 4 |
| Warm-up steps | 20 |
| Training pairs | 2,400 from the Charades-STA training split |
| Optimizer steps | 600 |
| Learning rate | 1e-5 |
| Seed | 0 |
| LoRA | r=16, alpha=32, dropout 0.05, on q/k/v/o and gate/up/down projections |
| Video input | 2 fps, at most 768 frames |
| Prompt template | timer1_default |
| Trainer | code/30_grpo_plus_placebo.py |
Anonymous. Accuracy Is Not Evidence: Measuring and Training Evidence Dependence in Video Temporal Grounding. ICLR 2027 submission, under review.
Qwen/Qwen2.5-VL-7B-Instruct