README.md

metadata

pretty_name: complex-func-bench agent trajectories
tags:
  - agent
  - function-calling
  - tool-use
  - llm-trajectories

AgentSuite/complex-func-bench-trajectories

Per-model agent trajectory data for complex-func-bench (public release).

• Models: 30
• Tasks per model: 1,000
• One file per model: {model}.jsonl, one JSON object per line.
• Fields: model_path, user_model_path, benchmark_name, task_name, sampling_params, user_sampling_params, messages, eval_result, meta.
• sampling_params reflect each benchmark's own implementation; values the benchmark leaves unset are recorded as null (provider default).

Models

• DeepSeek-R1-0528
• DeepSeek-V3-0324
• DeepSeek-V3.2-Exp-thinking-off
• DeepSeek-V3.2-Exp-thinking-on
• Kimi-K2-Instruct
• Kimi-K2-Instruct-0905
• Qwen3-235B-A22B-FP8
• Qwen3-235B-A22B-Instruct-2507-FP8
• Qwen3-235B-A22B-Thinking-2507-FP8
• Qwen3-Coder-480B-A35B-Instruct-FP8
• claude-4-opus-thinking-off
• claude-4-opus-thinking-on-10k
• claude-4-sonnet-thinking-off
• claude-4-sonnet-thinking-on-10k
• claude-4.5-sonnet-thinking-off
• claude-4.5-sonnet-thinking-on-10k
• gemini-2.5-flash-thinking-off
• gemini-2.5-flash-thinking-on
• gemini-2.5-pro-thinking-on
• gpt-4.1
• gpt-4.1-mini
• gpt-4.1-nano
• gpt-4o-20240806
• gpt-4o-mini
• gpt-5
• gpt-5-nano
• gpt-oss-120b
• gpt-oss-20b
• o3-high
• o4-mini-high