Usage
# pip install "transformers==4.51.3" "peft==0.20.0" qwen-vl-utilsimport torchfrom transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessorfrom 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-pc-a1-b003-s0")
The evaluation and audit scripts of the code repository take --model <base> --adapter <this repository or folder>.
Training
The placebo-controlled objective of Section 4 of the paper. It keeps the answer likelihood, weighted by α, and adds two margins for each edit operator: the annotated answer must be more likely on the clean video, and on a video with a matched placebo edit outside the event, than on the video with the event edited. A hinge keeps the drop under the placebo edit within a bound set by the reference model.
Table with columns: Setting, Value| Setting | Value |
|---|
| Operator weights | Group-DRO |
| α | 1.0 |
| β | 0.03 |
| Placebo bound | placebo drop ≤ 1.25 × reference drop + 0.10 |
| Training pairs | 2,400 from the Charades-STA training split |
| Optimizer steps | 300 |
| Gradient accumulation | 8 |
| Learning rate | 5e-6 |
| Seed | 0 |
| LoRA | r=16, alpha=32, dropout 0.05, on q/k/v/o and gate/up/down projections |
| Edit operators | swap, mask, blur, tshift, shuffle (two per example) |
| Video input | 2 fps, at most 768 frames |
| Prompt template | timer1_default |
| Trainer | code/27_papo_train_qcdro_v3seln.py |
Retrained model compared with the original
Both adapters were audited against Qwen2.5-VL-7B-Instruct on the first 200 Charades-STA test questions, with the same code under transformers 4.51.3. Γ, the paper's measure of evidence dependence, is in nats per token over the questions both models localise (n), with 95% video-clustered bootstrap intervals. R@0.5 is measured on all 3,720 test questions at 2 fps with greedy decoding, paired by question.
Table with columns: Retrained (this repository), Original run | Retrained (this repository) | Original run |
|---|
| Γ, mask | +0.85 [0.63, 1.05], n=105 | +0.82, n=105 |
| Γ, swap | +0.63 [0.48, 0.77], n=105 | – |
| R@0.5 change against the base | +5.9 points [4.4, 7.4] (56.6% vs 50.7%) | – |
| Answers with a parsable interval | 92.5% | – |
The original-run values are from Table 6 of the paper.
Citation
Anonymous. Accuracy Is Not Evidence: Measuring and Training Evidence Dependence in Video Temporal Grounding. ICLR 2027 submission, under review.