LesenForschungRadarAnlageframework
Anmelden / Registrieren
← Zurück zur Lektüre
NewsThema · FLEXKurzfassung2 Min. LesezeitDisconfirmAI Redaktion

MIT and NYU Researchers Develop FLEX System Using GPT-4o to Infer General Robot Preferences from Short Commands

Researchers at MIT and New York University developed FLEX, a robot learning system that uses OpenAI's GPT-4o to interpret short natural-language corrections as broader environmental rules. This allows robots to generalize safety constraints (e.g., avoiding all fragile objects after one warning about a mug) rather than just following literal commands, significantly improving intent understanding in motion planning.

2026-11-11 17:09 GMT+8

Researchers at the Massachusetts Institute of Technology (MIT) and New York University have developed an AI-assisted robot learning system called FLEX (Flexible Feature Editing Pipeline). The system addresses a key limitation in conventional robot motion planning: while a robot may understand a specific command, it often fails to recognize the underlying reason or generalize the constraint to similar situations. For example, if told not to spill coffee on a laptop, a standard robot might not avoid spilling on other nearby electronic devices unless explicitly instructed.

FLEX modifies the internal representation a robot uses when planning its movements by employing three operations: Add, Generalize, and Mask. It uses OpenAI’s GPT-4o language model to identify relationships between objects and interpret the intentions behind human corrections. If a robot carrying a bowling ball is warned not to drop it on a mug, the system infers that fragile objects should be avoided generally, rather than treating the mug as the only potential hazard. Similarly, a warning about carrying a lit candle near paper can be generalized into a preference for avoiding flammable materials.

The research was led by Andreea Bobu of MIT’s Computer Science and Artificial Intelligence Laboratory, with Audrey Lee, Nathaniel Dennler and Ilia Sucholutsky. The team evaluated the approach in two-dimensional navigation environments and simulated tabletop manipulation tasks involving a seven-degree-of-freedom Franka robotic arm. In sequential experiments, FLEX maintained higher normalized reward scores than approaches restricted to adding or removing features, which could accumulate unnecessary constraints or fail to generalize earlier instructions.

Anmelden / Registrieren
LesenForschungRadarAnlageframework

A separate study involved 12 participants who watched simulated robot movements and entered corrections in natural language. Participants rated FLEX more highly than comparison methods for understanding intent, with statistically significant differences reported for several measures. However, the distinction between simulation and physical deployment remains important. The participant study used simulated robot behavior, with real-robot demonstrations incorporated into offline evaluation. The researchers have not yet demonstrated a complete autonomous household robot operating under these conditions in real-time.

Quellen

  • eenewseurope.comeeNews Europe↗
Thema

FLEX

【背景】麻省理工学院(MIT)与纽约大学团队开发了FLEX系统,旨在解决机器人仅执行字面指令而忽略用户意图的痛点。该系统利用GPT-4o大语言模型,通过“添加”、“泛化”和“掩码”三种操作,将人类简短的自然语言纠正转化为更广泛的环境规则。例…

3 Schritte1 BeiträgeAktualisiert 2026-11-15 21:19

Wie es dazu kam

  1. 2026-10-20论文《Flexible Language-Guided Representation Editing for Adaptive Robot Behavior》提交至arXiv。arxiv.org ↗
  2. 2026-11-09Unite.AI报道MIT宣布FLEX研究成果,介绍其通过自然语言纠正优化机器人行为的能力。unite.ai ↗
  3. 2026-11-11eeNews Europe发布文章,详细解析FLEX如何利用GPT-4o从简短指令中推断通用偏好。eenewseurope.com ↗

Was noch ungeklärt ist

  • FLEX在真实物理环境中的闭环控制表现如何?目前主要依赖模拟轨迹和离线演示。
  • 对于非共享的、高度个性化的用户偏好,语言模型的常识推理是否足以准确泛化?
  • 项目代码和数据集何时正式公开?目前项目页面仍标记为“进行中”。
Vollständiges Thema lesen →
Recherche fortsetzen →