Update VLAC-Cut model card links and quick start

README.md

@@ -16,7 +16,7 @@ license: other
 
 **Video-based task progress estimation model for robot manipulation.**
 
-[Model](https://huggingface.co/InternRobotics/VLAC-Cut) · [Code](https://github.com/InternRobotics/VLAC-Cut) · [Paper](<PAPER_URL>) · [VPB Benchmark](https://huggingface.co/datasets/InternRobotics/VLAC-Cut-Benchmark)
+[Paper](https://arxiv.org/abs/2607.09776) · [Code](https://github.com/InternRobotics/VLAC-cut) · [Model](https://huggingface.co/InternRobotics/VLAC-Cut) · [Benchmark](https://huggingface.co/datasets/InternRobotics/VLAC-Cut-Benchmark)
 
 VLAC-Cut predicts how a robot task progresses over time from a natural-language instruction and an execution video. Instead of assuming that progress always increases with time, the model captures non-monotonic execution patterns including advancement, stagnation, regression, and recovery.
 
@@ -24,10 +24,9 @@ The model is designed for long-horizon robot manipulation videos containing fail
 
 ## Highlights
 
-- **Video-level progress understanding:** reasons over robot videos rather than isolated images or image pairs, and cuts the video into good segments and bad segments.
+- **Video-level progress understanding:** reasons over temporally ordered robot videos rather than isolated images or image pairs.
 - **Non-monotonic process modeling:** recognizes advancement, stagnation, regression, and recovery.
-- **General understanding capability:** General understanding capability, enabling zero-shot generalization across tasks, scenes, and viewpoints.
-- **Coarse/Fine-grained flexible:** The ability to adjust the frequency of video understanding, achieving progress analysis from coarse-grained to fine-grained.
+- **Generative output:** produces timestamped progress predictions through the standard autoregressive language-model interface, without a task-specific regression head.
 
 ## Model at a Glance
 
@@ -36,13 +35,13 @@ The model is designed for long-horizon robot manipulation videos containing fail
 | Base model | `Qwen/Qwen3-VL-30B-A3B-Instruct` |
 | Input | Task instruction, optional task plan, and sampled video frames |
 | Output | Timestamped task progress |
-| flexiable video sampling rate | `2 Hz~20Hz`(Default: `2.0`) |
+| Default video sampling rate | `2 Hz` |
 
 ## Model Design
 
 VLAC-Cut is fine-tuned from Qwen3-VL-30B-A3B-Instruct and retains its original multimodal generation interface. The model receives a task instruction, an optional task plan, and sampled video frames, then generates task progress at different timestamps as text.
 
-The training data also includes supervision for robot behavior description, failure analysis, and correction planning, helping the model understand realistic robot execution processes.
+The training data also include supervision for robot behavior description, failure analysis, and correction planning, helping the model understand realistic robot execution processes.
 
 This repository contains the model checkpoint and loading assets only. Inference scripts, examples, and VPB evaluation code are maintained in the GitHub repository.
 
@@ -67,15 +66,15 @@ model.eval()
 Run progress inference on a local video using the GitHub code:
 
 ```bash
-git clone https://github.com/InternRobotics/VLAC-Cut
-cd VLAC-Cut
+git clone https://github.com/InternRobotics/VLAC-cut
+cd VLAC-cut
 
 python scripts/run_example.py \
   --model-path InternRobotics/VLAC-Cut \
-  --video-path examples/example_01/episode.mp4 \
+  --video-path <path-to-video.mp4> \
   --task-instruction "<natural-language task instruction>" \
   --task-plan $'<optional step-by-step task plan>' \
-  --output-jsonl outputs/example_01.jsonl
+  --output-jsonl <path-to-output.jsonl>
 ```
 
 Main arguments:
@@ -100,24 +99,10 @@ Render a prediction preview video:
 
 ```bash
 python scripts/utils/render_prediction_video.py \
-  --input-jsonl outputs/example_01.jsonl \
-  --output-video outputs/example_01_pred_progress.mp4
+  --input-jsonl <path-to-output.jsonl> \
+  --output-video <path-to-preview.mp4>
 ```
 
-## Evaluation on VPB
-
-VLAC-Cut is evaluated on the Video Progress Benchmark. The benchmark data are released at `InternRobotics/VLAC-Cut-Benchmark`, and the evaluator is provided in the GitHub repository:
-
-```bash
-python scripts/evaluate_vpb_predictions.py \
-  --benchmark-root /path/to/VLAC-Cut-Benchmark/splits \
-  --predictions predictions/vlac_cut_predictions.jsonl \
-  --out-json reports/vlac_cut_vpb_eval.json \
-  --out-md reports/vlac_cut_vpb_eval.md
-```
-
-Report final VPB numbers here before public release.
-
 ## Citation
 
 Please replace the placeholder below with the final paper BibTeX before public release.