RUCAIBox/HaluEval

Issues

• The issue with newline characters in the JSON file of your instructions for generating multi-turn hallucination responses.

  #22 opened 3 months ago by wizaaaard
  1

• Why are responses that ChatGPT refuses to answer considered as hallucinations?

  #21 opened 10 months ago by JacksonWuxs
  1

• I added HaluEval to lm-eval-harness, can you please double check?

  #18 opened a year ago by pminervini
  3

• A minor potential problem

  #15 opened a year ago by Vicent0205
  1

• 非常赞的工作，请问后续有针对中文场景开源幻觉评估数据的计划吗

  #12 opened a year ago by wen2cheng
  1

• How to retrieve relevant knowledge

  #13 opened a year ago by zeta-zl
  1

• Open Source License?

  #14 opened a year ago by ilanazim
  1

• Is it a good choice to use Accuracy in evaluating the hallucination detection/recognition performance?

  #20 opened a year ago by JacksonWuxs
  1

• JSONL format?

  #19 opened a year ago by forrestbao
  0

• In the QA setting, the "knowledge" part of the question is not being passed to the model

  #17 opened a year ago by pminervini
  3

• Evaluation with OpenCompass

  #11 opened a year ago by tonysy
  0

• How did you determine that the max token length for truncated message was 2033?

  #9 opened a year ago by Whisht
  2