README.md

metadata

task_categories:
  - robotics
tags:
  - vla
  - semantic-grounding
  - benchmark
license: other

RoboSemanticBench

Paper | GitHub

RoboSemanticBench (RSB) is an embodied benchmark for diagnosing semantic grounding in action prediction for vision-language-action (VLA) models. In each episode, the robot receives a multiple-choice math or general-knowledge question, observes candidate answer blocks, and must pick the physical block corresponding to the correct answer. This setup isolates the model's semantic decision-making from its low-level manipulation capability.

Benchmark Suites

The benchmark consists of six suites with 4-choice and 10-choice variants:

Suite	Task name	Choices	Semantic source
RSB-Math-4	rsb_math	4	Procedural arithmetic
RSB-Math-10	rsb_math_10blocks	10	Procedural arithmetic
RSB-HardMath-4	rsb_hardmath	4	GSM8K
RSB-HardMath-10	rsb_hardmath_10blocks	10	GSM8K
RSB-General-4	rsb_general	4	MMLU-style QA
RSB-General-10	rsb_general_10blocks	10	MMLU-style QA

Metrics

RSB reports three diagnostic metrics:

• 🎯 Task Success Rate (TSR): The episode succeeds only when the robot selects the semantically correct answer block and places it in the answer zone.
• ✋ Grasp Success Rate (GSR): The robot successfully grasps any candidate answer block, regardless of whether it is correct.
• 🧭 Normalized Semantic Grounding (nSG): Semantic target selection conditioned on successful grasping, where 0 corresponds to random target selection.

Sample Usage

Data Collection

To collect trajectories for a specific task (e.g., RSB-Math-4):

# RSB-Math-4
bash collect_data.sh rsb_math rsb_math_train_500 0 --skip_instructions

Policy Evaluation

To evaluate a policy using the simulation evaluation entry:

python script/eval_policy.py --config policy/Your_Policy/deploy_policy.yml \
  --overrides \
  --task_name rsb_general_10blocks \
  --task_config rsb_general_10blocks_test_500 \
  --ckpt_setting your_checkpoint_name \
  --seed 0 \
  --policy_name Your_Policy

Citation

@misc{RoboSemanticBench,
      title={RoboSemanticBench: Diagnosing Semantic Grounding in Action Prediction for VLA Models}, 
      author={Bin Yu and Yao Zhang and Haishan Liu and Shijie Lian and Yuliang Wei and Xiaopeng Lin and Zhaolong Shen and Changti Wu and Ruina Hu and Bailing Wang and Cong Huang and Kai Chen},
      year={2026},
      eprint={2606.02277},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2606.02277}, 
}