Tool use / function calling

Most-superseded baselines

Ranked by how many distinct papers critique or beat each method. These are the standard baselines newer work routinely measures against.

1. 1

   ReAct· ReAct

   ReAct: Synergizing Reasoning and Acting in Language Models

   6 papers critique it · 4 beat it on benchmarks

2. 2

   ToolLLM· ToolLLM

   ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs

   4 papers critique it · 3 beat it on benchmarks

3. 3

   ToolACE· ReAct

   ToolACE: Winning the Points of LLM Function Calling

   2 papers critique it · 4 beat it on benchmarks

4. 4

   API-Bank· ReAct

   API-Bank: A Comprehensive Benchmark for Tool-Augmented LLMs

   5 papers critique it · 0 beat it on benchmarks

5. 5

   Gorilla· ToolLLM

   Gorilla: Large Language Model Connected with Massive APIs

   3 papers critique it · 1 beat it on benchmarks

6. 6

   GRPO· GRPO

   Reinforcement Learning Outperforms Supervised Fine-Tuning: A Case Study on Audio Question Answering

   2 papers critique it · 1 beat it on benchmarks

7. 7

   StepTool· StepTool

   2 papers critique it · 1 beat it on benchmarks

8. 8

   Toolformer· Toolformer

   Toolformer: Language Models Can Teach Themselves to Use Tools

   2 papers critique it · 1 beat it on benchmarks

9. 9

   APIGen· ReAct

   APIGen: Automated Pipeline for Generating Verifiable and Diverse Function-Calling Datasets

   3 papers critique it · 0 beat it on benchmarks

10. 10

    xLAM· ReAct

    xLAM: A Family of Large Action Models to Empower AI Agent Systems

    0 papers critique it · 2 beat it on benchmarks

11. 11

    ART· ART

    ART: Automatic multi-step reasoning and tool-use for large language models

    1 papers critique it · 1 beat it on benchmarks

12. 12

    ExpeL· ART

    ExpeL: LLM Agents Are Experiential Learners

    1 papers critique it · 1 beat it on benchmarks