Instructions to use tsinghua-sigs-robot-lab/VeriLoop-E2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsinghua-sigs-robot-lab/VeriLoop-E2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="tsinghua-sigs-robot-lab/VeriLoop-E2")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("tsinghua-sigs-robot-lab/VeriLoop-E2")
model = AutoModelForMultimodalLM.from_pretrained("tsinghua-sigs-robot-lab/VeriLoop-E2", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tsinghua-sigs-robot-lab/VeriLoop-E2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsinghua-sigs-robot-lab/VeriLoop-E2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsinghua-sigs-robot-lab/VeriLoop-E2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsinghua-sigs-robot-lab/VeriLoop-E2

SGLang

How to use tsinghua-sigs-robot-lab/VeriLoop-E2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "tsinghua-sigs-robot-lab/VeriLoop-E2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsinghua-sigs-robot-lab/VeriLoop-E2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "tsinghua-sigs-robot-lab/VeriLoop-E2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsinghua-sigs-robot-lab/VeriLoop-E2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsinghua-sigs-robot-lab/VeriLoop-E2 with Docker Model Runner:
```
docker model run hf.co/tsinghua-sigs-robot-lab/VeriLoop-E2
```

ConorWang commited on 11 days ago

Commit

a5c1ca2

verified ·

1 Parent(s): 483fd22

Add VeriLoop E2 structured evaluation results

Browse files

Files changed (9) hide show

.eval_results/README.md +42 -0
.eval_results/aime-2026.yaml +7 -0
.eval_results/apex-2025.yaml +7 -0
.eval_results/deep-swe.yaml +8 -0
.eval_results/gpqa.yaml +7 -0
.eval_results/swe-bench-pro.yaml +7 -0
.eval_results/terminal-bench-2.1.yaml +7 -0
.eval_results/terminal-bench-3.0.yaml +7 -0
.eval_results/terminal-bench-4.0.yaml +7 -0

.eval_results/README.md ADDED Viewed

	@@ -0,0 +1,42 @@

+# VeriLoop E2 Evaluation Results
+This directory contains the structured Hugging Face evaluation-result descriptors for **VeriLoop E2**. Each YAML file records a benchmark score together with the corresponding benchmark registration metadata and a direct provenance link to the public evaluation evidence released in the companion dataset:
+**Evaluation evidence:**
+https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence
+The files in this directory are intended for Hugging Face model-page evaluation display and, where the benchmark is registered as a Hugging Face Native Benchmark, automatic leaderboard aggregation.
+## Registered Evaluation Results
+| File | Benchmark | VeriLoop E2 | Public Evidence |
+|---|---|---:|---|
+| `aime-2026.yaml` | AIME 2026 | **98.3%** | `VeriLoop-E2-Evaluation-Evidence/aime-2026` |
+| `apex-2025.yaml` | MathArena Apex 2025 | **89.6%** | `VeriLoop-E2-Evaluation-Evidence/apex-2025` |
+| `deep-swe.yaml` | DeepSWE v1.1 | **64.6%** | `VeriLoop-E2-Evaluation-Evidence/deepswe-1.1` |
+| `gpqa.yaml` | GPQA Diamond | **93.94%** | `VeriLoop-E2-Evaluation-Evidence/gpqa-diamond` |
+| `swe-bench-pro.yaml` | SWE-bench Pro | **76.2%** | `VeriLoop-E2-Evaluation-Evidence/swe-bench-pro` |
+| `terminal-bench-2.1.yaml` | Terminal-Bench 2.1 | **88.8%** | `VeriLoop-E2-Evaluation-Evidence/terminal-bench-2.1` |
+| `terminal-bench-3.0.yaml` | Terminal-Bench 3.0 | **29.7%** | `VeriLoop-E2-Evaluation-Evidence/terminal-bench-3.0` |
+| `terminal-bench-4.0.yaml` | Terminal-Bench 4.0 | **37.9%** | `VeriLoop-E2-Evaluation-Evidence/terminal-bench-4.0` |
+## SWE-Marathon v1.1
+**SWE-Marathon v1.1 is intentionally not represented by an `.eval_results/*.yaml` file in this directory.**
+VeriLoop E2 achieved **45.0%** on SWE-Marathon v1.1. The result is reported on the VeriLoop E2 model page and is backed by the corresponding public evaluation evidence. At the time of this release, SWE-Marathon v1.1 is not exposed through a verified Hugging Face Native Benchmark registration with a stable `task_id` suitable for automatic `.eval_results` leaderboard aggregation.
+Accordingly, no synthetic or guessed benchmark identifier is used here. This keeps the machine-readable evaluation metadata strictly aligned with benchmark registrations that can be independently verified.
+## Provenance Policy
+Every score represented in this directory is paired with a benchmark-specific evidence path in the public `VeriLoop-E2-Evaluation-Evidence` dataset. The model repository serves as the structured score-registration layer, while the evidence dataset serves as the public audit layer for task-level evaluation artifacts.
+Reported scores are not reconstructed from the YAML files themselves; the YAML files reference already-released evaluation evidence. Any future correction or benchmark-registration update should preserve this separation between score metadata and underlying evidence.
+## Model
+**Model:** VeriLoop E2
+**Organization:** Tsinghua SIGS Robot Lab
+**Model repository:** https://huggingface.co/tsinghua-sigs-robot-lab/VeriLoop-E2
+**Evaluation evidence:** https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence

.eval_results/aime-2026.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: MathArena/aime_2026
+    task_id: MathArena/aime_2026
+  value: 98.3
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/aime-2026
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/apex-2025.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: MathArena/apex_2025
+    task_id: MathArena/apex_2025
+  value: 89.6
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/apex-2025
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/deep-swe.yaml ADDED Viewed

	@@ -0,0 +1,8 @@

+- dataset:
+    id: datacurve/deep-swe
+    task_id: deep_swe
+  value: 64.6
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/deepswe-1.1
+    name: VeriLoop E2 Evaluation Evidence
+  notes: DeepSWE v1.1.

.eval_results/gpqa.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: Idavidrein/gpqa
+    task_id: diamond
+  value: 93.94
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/gpqa-diamond
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/swe-bench-pro.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: ScaleAI/SWE-bench_Pro
+    task_id: SWE_Bench_Pro
+  value: 76.2
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/swe-bench-pro
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/terminal-bench-2.1.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: harborframework/terminal-bench-2.1
+    task_id: terminalbench_2_1
+  value: 88.8
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/terminal-bench-2.1
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/terminal-bench-3.0.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: harborframework/terminal-bench
+    task_id: terminalbench_3
+  value: 29.7
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/terminal-bench-3.0
+    name: VeriLoop E2 Evaluation Evidence

.eval_results/terminal-bench-4.0.yaml ADDED Viewed

	@@ -0,0 +1,7 @@

+- dataset:
+    id: harborframework/terminal-bench
+    task_id: terminalbench_4
+  value: 37.9
+  source:
+    url: https://huggingface.co/datasets/tsinghua-sigs-robot-lab/VeriLoop-E2-Evaluation-Evidence/tree/main/terminal-bench-4.0
+    name: VeriLoop E2 Evaluation Evidence