[BFCL] GPT-5 performs poorly on bfcl — seeking validation from community #1157
Description
Describe the issue
I recently evaluated GPT-5 on the benchmark and observed significantly worse performance than expected, Achieving results substantially below comparable models (e.g., GPT-4 or other baselines).
ID datapoint
What is the issue
Has anyone else encountered similar issues? I’m curious whether this might be related to:
- Specific evaluation settings or prompts
- Domain-specific limitations of the model
- Potential implementation or deployment quirks
If you’ve run similar experiments or have insights, please share your findings or suggestions below.
Proposed Changes
Additional context
I'm looking forward to seeing the official evaluation results from on this leaderboard to better understand the model's capabilities. Thank you!