Issues

• compute_metrics() scores a compiling generation as a full pass when the problem has zero test cases (741 TACO train problems affected)

  FlagOpen/TACO#21 · Filter by author shaurya416 opened on Oct 1, 2026

• How do you evaluate generated code for questions that accept multiple possible outputs for the same input

  FlagOpen/TACO#20 · Filter by author zb1439 opened on Feb 1, 2026

• solutions缺失

  FlagOpen/TACO#19 · Filter by author bPARADOXc opened on Aug 29, 2025

• Test data contains inputs outside problem constraints in taco_generation_example dataset

  FlagOpen/TACO#18 · Filter by author Mr-zhouyi opened on Aug 13, 2025

• compute_metric.py评估问题

  FlagOpen/TACO#17 · Filter by author jiangyuan1018 opened on Mar 23, 2025
  1

• Questions about the generation of the test cases

  FlagOpen/TACO#16 · Filter by author twelfth-star opened on Feb 21, 2025

• Problem Regarding Evaluation

  FlagOpen/TACO#15 · Filter by author chrisliu298 opened on Oct 25, 2024
  3

• Taco数据集能够用于其他模型微调，比如qwen，或者是和llama factory框架适配

  FlagOpen/TACO#14 · Filter by author zhenzhenzhizhi opened on Oct 24, 2024