Research map / Vision-language and multimodal / Robot action and navigation
E-TTS: A New Embodied Test-Time Scaling Framework for Robotic Manipulation
Wen Ye, Peiyan Li, Tingyu Yuan, Yuan Xu, Xiangnan Wu, Chaoyang Zhao, Jing Liu, Nianfeng Liu, Yan Huang, Liang Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add E-TTS | E-CoT · SimplerEnv WidowX | average success rate: 6.67 -> 39.81 | Improved |
| add E-TTS instead of RoboMonkey | E-CoT · SimplerEnv WidowX | average success rate: 26.38 -> 39.81 | Improved |
| add feedback generation (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 25.04 -> 39.81 | Improved |
| add reasoning scaling to action scaling (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 26.39 -> 39.81 | Improved |
| add action scaling to reasoning scaling (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 30.56 -> 39.81 | Improved |
| add joint reasoning-action scoring (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 31.13 -> 39.81 | Improved |
| add history buffer (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 31.94 -> 39.81 | Improved |
| add epsilon-greedy exploration (ablation) | E-TTS · SimplerEnv WidowX | average success rate: 36.11 -> 39.81 | Improved |
| increase history length K from 1 to 10 (ablation) | E-TTS · SimplerEnv WidowX | success rate: 37.50 -> 58.33 | Improved |
| increase history length K from 10 to 20 (ablation) | E-TTS · SimplerEnv WidowX | success rate: 58.33 -> 45.83 | Hurt |
| increase threshold eta from 0.1 to 0.5 (ablation) | E-TTS · SimplerEnv WidowX | success rate: 70.83 -> 33.33 | Hurt |
| increase epsilon from 0.1 to 0.7 (ablation) | E-TTS · SimplerEnv WidowX | success rate: 54.17 -> 33.33 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Embodied-R1 — Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation (ICLR 2026) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models (NeurIPS 2026 · similarity 0.90)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.89)
- RoboCerebra: A Large-scale Benchmark for Long-horizon Robotic Manipulation Evaluation (NeurIPS 2025 · similarity 0.89)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 2026 · similarity 0.89)
- Learning from Trials and Errors: Reflective Test-Time Planning for Embodied LLMs (NeurIPS 2026 · similarity 0.89)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
