Syntax Error-Free and Generalizable Tool Use for LLMs: Appendix

cover
2 Jun 2024

Authors:

(1) Kexun Zhang, UC Santa Barbara and Equal contribution;

(2) Hongqiao Chen, Northwood High School and Equal contribution;

(3) Lei Li, Carnegie Mellon University;

(4) William Yang Wang,UC Santa Barbara.

A. APPENDIX

A.1 PSEUDO-CODE OF THE DECODING ALGORITHM

A.2 GENERALIZING TOOLKENGPT TO UNSEEN NEW TOOLS

Figure 6: Once a tool call begins, TOOLDEC injects a special prompt (blue text) into context to generate the tool name.

In this section, we show examples of TOOLDEC preventing tool-related errors on various baselines. Baselines are displayed in the left column and TOOLDEC is showed on the right.

Figure 7: TOOLDEC can prevent function name error, function argument error, and invalid ReAct syntax on ToolLLM.

Figure 8: TOOLDEC can eliminate the common tool-related errors for fine-tuned models.

A.4 EXAMPLES OF KAMEL RELATIONS

Table 6: Examples of KAMEL Relations

This paper is available on arxiv under CC 4.0 DEED license.