Logo

GitLab

Sign in

朱子纯 / xlyNewAi

  • Back to Dashboard
  • Project
  • Activity
  • Files
  • Commits
  • Network
  • Graphs
  • Milestones
  • Issues 0
  • Merge Requests 0
  • Labels
  • Wiki
  • Snippets
  • Commits 2
  • Compare
  • Branches 2
  • Tags 0
  • xlyNewAi
  • bench
28 Jul, 2026
3 commits
  • skills: ASCII filenames (unzip mangles CJK names in deploy chain), name moved in… ...
    ee583b89
    …to file line 1; bench: simulate production anti-fab retry guard
    zichun authored
    2026-07-28 13:40:01 +0800  
    Browse Code »
  • P2: skill-ReAct — drop intent gate and deterministic pipelines ...
    b7543403
    - single ReAct entry for all free text: 7 fixed tools (useSkill + 6);
      flow knowledge lives in 5 skill files (resources/skills/*.md, first
      line = brief for prompt index; xly.skills.dir overrides for hot reload)
    - useSkill returns skill body and logs skill_active; projection pins the
      active skill full text on the flow card until confirm/cancel/new skill
    - system prompt from template prompts/system.txt (domain map + skill
      index); bench reads the same files to avoid replica drift
    - delete IntentService/Intent/SlotFillService/LlmJsonClient/StateService/
      OkHttpUtil and route()/handleCreate/handleWrite/deriveWriteAction/
      ground/withState; dimension-split regex migrated into FormCollectTool
      prefill assist (model provably fails dimension splitting)
    - anti-fabrication guards now unconditional (zero-tools+digits retry,
      WRITE_CLAIM check); invariant layer and button endpoints untouched
    - bench_ext: --arch new (same-source prompt/skills) and --fifty
      (bench50 cases judged on full trajectory for both arches)
    zichun authored
    2026-07-28 13:34:58 +0800  
    Browse Code »
  • bench: freeze bench50_v2 baseline; add trajectory bench (multi-turn/disambiguati… ...
    8bd01fbd
    …on/form-continuation); old-arch baseline 9/12
    zichun authored
    2026-07-28 13:14:14 +0800  
    Browse Code »