TRACE Qwen2.5-VL 7B
TRACE Qwen2.5-VL 7B is a GRPO checkpoint derived from
Qwen/Qwen2.5-VL-7B-Instruct
and trained on 64,000 grounded visual-reasoning examples spanning 1,000 tasks
from
maveryn/trace. This
repository contains the merged step-500 inference checkpoint.
Paper · Project page · GitHub · Collection · Training configuration · Evaluation suite · Run artifacts
Training and provenance
| Field | Value |
|---|---|
| Base model | Qwen/Qwen2.5-VL-7B-Instruct@cc594898137f460bfe9f0759e9844b3ce807cfb5 |
| Dataset | maveryn/trace@4e5b54361360296a855542b40cfd8b7f81b355fe |
| Method | GRPO, 500 steps, global/rollout batch 128, 8 responses per prompt |
| Optimization | AdamW, learning rate 1e-6, constant schedule, KL disabled |
| Reward | 0.95 × exact JSON answer + 0.05 × valid JSON format; annotation reward 0 |
| Reference run | 8 NVIDIA H100 80GB GPUs |
| Released artifact | Merged step-500 inference weights |
The released training profile reads prompt_answer, scores answer_gt, and
does not use the advisory trace_supervision_mode column. The consumed fields,
embedded image bytes, and row order in the current dataset release were
verified identical to the original training input in the public
equivalence receipt.
The canonical output ends with {"answer": ...}. Exact hashes, source
revision, and run provenance are recorded in
trace_training_provenance.json and
.trace_model_revision.json. The repository
does not include optimizer, scheduler, trainer, or FSDP state for continuation.
Evaluation
trace_eval_v1 evaluates 24 external benchmarks and 32,805 examples per model
and decoding seed. Scores below are the unweighted macro mean of the 24
benchmark percentages, reported as mean ± sample standard deviation across
seeds 42, 43, and 44.
| Model | Overall score |
|---|---|
| Qwen2.5-VL-7B-Instruct | 47.93 ± 0.30 |
| TRACE Qwen2.5-VL 7B | 51.99 ± 0.17 |
| VERO Qwen2.5-VL 7B | 52.22 ± 0.07 |
| Game-RL Qwen2.5-VL 7B | 48.05 ± 0.45 |
| Sphinx Qwen2.5-VL 7B | 49.34 ± 0.15 |
| PCGRPO Qwen2.5-VL 7B | 48.80 ± 0.28 |
| Paired improvement | +4.06 ± 0.41 |
Per-benchmark scores, model revisions, aggregation, and benchmark provenance
are available in the public
results.json
and evaluation documentation.
Usage
from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration
model_id = "maveryn/trace-qwen2.5-vl-7b"
revision = "4d0f1ae8ee25022058090dbdbff61957ece7331d"
image_url = "https://raw.githubusercontent.com/maveryn/trace/main/docs/assets/paper-domain-montage/trace-paper-domain-montage.png"
processor = AutoProcessor.from_pretrained(model_id, revision=revision)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
model_id,
revision=revision,
dtype="auto",
device_map="auto",
)
messages = [
{
"role": "user",
"content": [
{"type": "image", "url": image_url},
{"type": "text", "text": "How many visual domains are shown with example images? Respond with only a JSON object using the key \"answer\"."},
],
}
]
inputs = processor.apply_chat_template(
messages,
tokenize=True,
add_generation_prompt=True,
return_dict=True,
return_tensors="pt",
).to(model.device)
output_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids = [
output[len(prompt) :] for prompt, output in zip(inputs.input_ids, output_ids)
]
print(processor.batch_decode(generated_ids, skip_special_tokens=True)[0])
Verified reference answer
For the published montage used above, the expected output is:
{"answer": 11}
The value is backed by the committed
montage manifest
(layout.panel_count). This is reference ground truth, not a claim about a
particular decoding run; publish qualitative model outputs only with their
recorded inference settings.
The pinned revision is the checkpoint used by the canonical evaluation; later repository heads may update documentation without changing the model weights.
Intended use and limitations
This checkpoint is intended for research on multimodal reasoning and verifiable-reward post-training. It can produce incorrect answers, unsupported reasoning, or unreliable grounding. It was trained on synthetic tasks and has not been validated for safety-critical, medical, legal, or autonomous decision-making uses. Results depend on the recorded prompts, decoding, parsers, and scorers.
License
This checkpoint is released under Apache-2.0, consistent with the upstream base model.