LoRA adapter for Qwen2.5-VL-7B-Instruct, trained with supervised fine-tuning (SFT) on 4,800 training pairs, with its checkpoint at step 100.
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 26, 28, 29, 37 and 38 (Appendix F).
checkpoints/step-100/ is the same run after 100 of its 600 steps (800 pairs). Table 38 uses this checkpoint.
| Folder | Contents | transformers |
|---|---|---|
| root | final adapter | 5.x (tested 5.12.1, peft 0.19.1 or later) |
checkpoints/step-100/ |
checkpoint at step 100 | 5.x (tested 5.12.1, peft 0.19.1 or later) |
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.
# pip install "transformers>=5.12" "peft>=0.19.1" 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-sft-4800pairs-s0")
The evaluation and audit scripts of the code repository take --model <base> --adapter <this repository or folder>.
Supervised fine-tuning with the same trainer and data as the placebo-controlled models: the answer likelihood only, with the margin terms switched off.
| Setting | Value |
|---|---|
| Training pairs | 4,800 from the Charades-STA training split |
| Optimizer steps | 600 |
| 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 |
| Video input | 2 fps, at most 768 frames |
| Prompt template | timer1_default |
| Trainer | code/27_papo_train_qcdro_v3seln_r5.py |
Anonymous. Accuracy Is Not Evidence: Measuring and Training Evidence Dependence in Video Temporal Grounding. ICLR 2027 submission, under review.
adapter_model.safetensors 190 MBcheckpoints/step-100/adapter_model.safetensors 190 MBQwen/Qwen2.5-VL-7B-Instruct