Issues

• A bug in the PostToolUse hook causes excessive duplication of observations in the database, with each tool use creating 6-10 identical records instead of one. Despite efforts to clean up 51,140 duplicate entries and leave only 6,098 unique observations, the issue reoccurs immediately after new tool usage. Investigation confirms there is only one registered PostToolUse hook, and no additional configuration anomalies, indicating the problem likely lies in the worker-service.cjs file, which lacks deduplication logic in the observation creation pipeline. Evidence includes screenshots of identical entries and cleanup scripts identifying duplicate groups. The suggested fix involves introducing hash-based or time-window deduplication before inserting observations into the SQLite database to ensure each tool use results in a single, unique record.

  zai-org/ComplexFuncBench#12 · Filter by author Cyprianhenry opened on Feb 18, 2026
  1

• ## Code Review: bugfix/linger

  zai-org/ComplexFuncBench#11 · Filter by author Cyprianhenry opened on Dec 28, 2025

• http://xx.xx.xx.xx:8000/v1

  zai-org/ComplexFuncBench#10 · Filter by author Cyprianhenry opened on Dec 23, 2025

• Please add a license to this repo

  zai-org/ComplexFuncBench#8 · Filter by author nrafidi opened on Oct 27, 2025
  2

• Getting stuck at tool call in evaluation of Qwen3

  zai-org/ComplexFuncBench#7 · Filter by author Aliasgarsaifee opened on Sep 19, 2025
  1

• Ground truth response in the evaluation dataset

  zai-org/ComplexFuncBench#6 · Filter by author antonpibm opened on Sep 9, 2025
  3

• GPT-5 performs very bad on this benchmark

  zai-org/ComplexFuncBench#5 · Filter by author Huluwa-kong opened on Aug 15, 2025
  1

• GPTRunner discards text response when there is tool call

  zai-org/ComplexFuncBench#4 · Filter by author vantaa89 opened on Aug 14, 2025
  1