• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI
    Announcement

    X-Tree proposes a tree of reusable experience for AI agents

    Its creators say they build the tree without LLMs and use it as data, reward and context for training.

    Victor ZhongVZ
    Sitao Cheng @COLM 2026SC
    2 Sources, ,

    TLDR

    X-Tree’s creators argue that training agents on flat action sequences leaves reusable experience unused. They say they mine that experience into a tree without LLMs, then use the tree as data, reward and context for training.

    Combined views

    469

    2 Sources, first seen 7h ago

    Combined views

    469

    2 Sources, first seen 7h ago

    10 likes
    7h ago
    first seen 7h ago
    10 likes
    9 comments
    2 saves
    8 reposts
    9 comments
    2 saves
    8 reposts
    Featured Source

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    2 Sources

    Sitao Cheng @COLM 2026@TonyCheng990417Agents learn from flat action sequences (SFT, RLVR), so reusable experience go unused. We mine them into a tree without LLMs, and train on it as data, reward and context. X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization 🧵 🔗 https://sitaocheng.github.io/xtree/7h
    Victor Zhong@hllo_wrldRT @TonyCheng990417: Agents learn from flat action sequences (SFT, RLVR), so reusable experience go unused. We mine them into a tree wit…2h

    2 Sources

    Sitao Cheng @COLM 2026@TonyCheng990417Agents learn from flat action sequences (SFT, RLVR), so reusable experience go unused. We mine them into a tree without LLMs, and train on it as data, reward and context. X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization 🧵 🔗 https://sitaocheng.github.io/xtree/7h
    Victor Zhong@hllo_wrldRT @TonyCheng990417: Agents learn from flat action sequences (SFT, RLVR), so reusable experience go unused. We mine them into a tree wit…2h