Shanghai Jiao Tong University study on robot scene entropy vulnerability and Rob
Overview
RoboPrompt is a robot control framework proposed by a team from Shanghai Jiao Tong University, which allows users to intuitively guide robot actions through sparse inputs such as points, trajectories, or directional commands. The framework employs a two-stage design, translating human intent into action drafts and then refining their execution through a lower-level policy. Experiments show that RoboPrompt significantly improves the success rate of robot tasks and reduces the need for manual intervention.
From QbitAI — Robotics on WeChat
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · Reported success rate: 15.5 percent · Basis not reported · Differing source assertions
“以π0.5为例，在Insert Bread、Hang Cup和Push Ball三个真实任务上，2—3轮迭代后平均成功率提升 15.5% ，平均人工干预次数从 2.86次 降至 1.60次 ，减少 44.0% 。”
Exact source · revision 1qbitai
Reported quantity · Reported success rate: 21.3 percent · Basis not reported · Differing source assertions
“在Insert Bread任务上，跨Diffusion Policy、π0.5和FastWAM三种策略的平均成功率提升 21.3% ，干预次数从 2.60次 降至 0.47次 ，减少 81.9% 。”
Exact source · revision 1qbitai
Reported quantity · intervention count: 2.86 other · Basis not reported · Differing source assertions
“以π0.5为例，在Insert Bread、Hang Cup和Push Ball三个真实任务上，2—3轮迭代后平均成功率提升 15.5% ，平均人工干预次数从 2.86次 降至 1.60次 ，减少 44.0% 。”
Exact source · revision 1qbitai
Reported quantity · intervention count: 2.6 other · Basis not reported · Differing source assertions
“在Insert Bread任务上，跨Diffusion Policy、π0.5和FastWAM三种策略的平均成功率提升 21.3% ，干预次数从 2.60次 降至 0.47次 ，减少 81.9% 。”
Exact source · revision 1qbitai
Reported quantity · scene entropy: 3.782 other · Basis not reported · Differing source assertions
“研究团队专门计算了这个数据集的场景熵，达到了3.782比特，作为对比，业内常用的LIBERO基准里几个子任务集合的场景熵平均只有0.88比特。”
Exact source · revision 2Source owner not reported
Reported quantity · training tasks: 75 other · Basis not reported
“整个数据集一共设计了75种训练任务标签，归入六组场景配置，总共产生了3750条训练演示轨迹。”
Exact source · revision 2Source owner not reported
Reported quantity · scene entropy: 0.88 other · Basis not reported · Differing source assertions
“研究团队专门计算了这个数据集的场景熵，达到了3.782比特，作为对比，业内常用的LIBERO基准里几个子任务集合的场景熵平均只有0.88比特。”
Exact source · revision 2Source owner not reported
Evidence dependency changes (4)
- 2026-10-30T22:39:40.141Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:39:23.074Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:31:15.825Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T22:30:59.589Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
Developments
- 2026-10-18 17:04 UTC · 1 reportsShanghai Jiao Tong University study on robot scene entropy vulnerability and RobChinese robotics — Research news discovery：RoboFollow: A Benchmark for Evaluating Language Understanding in Robot Control
- 2026-10-18 10:25 UTC · 1 reportsRoboPrompt: Sparse Input Strategy Framework for Robot ControlQbitAI — Robotics on WeChat：Drawing Lines to Control Robots: Shanghai Jiao Tong University Proposes a Sparse Input Strategy Guidance Framework
Report timeline
Follow attributed reports and material updates.
- Chinese robotics — Research news discoverySignalRoboFollow: A Benchmark for Evaluating Language Understanding in Robot Control
A research team from Shanghai Jiao Tong University identifies a critical flaw in current robot control models: they rely on visual shortcuts rather than true language understanding. To address this, they introduce RoboFollow, a benchmark with high-scene-entropy scenarios that force models to depend on language for task resolution. The study shows that even top models fail under slight variations in scene layout or instruction phrasing, indicating a deeper issue in training paradigms.
- QbitAI — Robotics on WeChatSignalDrawing Lines to Control Robots: Shanghai Jiao Tong University Proposes a Sparse Input Strategy Guidance Framework
RoboPrompt is a robot control framework proposed by a team from Shanghai Jiao Tong University, which allows users to intuitively guide robot actions through sparse inputs such as points, trajectories, or directional commands. The framework employs a two-stage design, translating human intent into action drafts and then refining their execution through a lower-level policy. Experiments show that RoboPrompt significantly improves the success rate of robot tasks and reduces the need for manual intervention.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).