List your company
ResearchFormerly WizWand

Function Selection on TinyAgent

94.2Function Selection AccuracyDTDR-L
2.36826.20950.0573.891Dec 18, 2025
Updated 8mo ago

Evaluation results

MethodLinks
DTDR-LBackbone=GPT-4o2025.1294.2
DTDR-LBackbone=Qwen3 14B2025.1292.5
DTDR-LBackbone=Qwen3 8B2025.1289
Raw DemonstrationsBackbone=GPT-4o2025.1286.8
Dependency RetrievalBackbone=GPT-4o2025.1285.9
DTDR-CBackbone=GPT-4o2025.1285.9
DTDR-CBackbone=Qwen3 14B2025.1284.8
No ICLBackbone=GPT-4o2025.1284.7
DTDR-LBackbone=Qwen3 4B2025.1280.7
DTDR-CBackbone=Qwen3 8B2025.1280.4
Dependency RetrievalBackbone=Qwen3 14B2025.1279.5
DTDR-LBackbone=Qwen3 1.7B2025.1278.4
No ICLBackbone=Qwen3 14B2025.1278
Raw DemonstrationsBackbone=Qwen3 14B2025.1277.9
Dependency RetrievalBackbone=Qwen3 8B2025.1276.7
No ICLBackbone=Qwen3 8B2025.1271.3
DTDR-CBackbone=Qwen3 4B2025.1268.1
Raw DemonstrationsBackbone=Qwen3 8B2025.1267.8
Dynamic LR (DTDR-L)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1265.1
DTDR-CBackbone=Qwen3 1.7B2025.1264.1
Dependency RetrievalBackbone=Qwen3 4B2025.1262.7
Raw DemonstrationsBackbone=Qwen3 4B2025.1260.2
No ICLBackbone=Qwen3 4B2025.1259.6
DTDR-LBackbone=Gorilla V22025.1258.7
Learned RetrievalBackbone=GPT-4o2025.1253.2
Learned RetrievalBackbone=Qwen3 4B2025.1252.2
Dependency RetrievalBackbone=Qwen3 1.7B2025.1251.6
Learned RetrievalBackbone=Qwen3 14B2025.1251.1
Learned RetrievalBackbone=Qwen3 1.7B2025.1250.5
No ICLBackbone=Qwen3 1.7B2025.1248.1
Raw DemonstrationsBackbone=Qwen3 1.7B2025.1246.3
DTDR-LBackbone=Qwen3 0.6B2025.1246.1
Learned RetrievalBackbone=Qwen3 8B2025.1243.7
DTDR-CBackbone=Gorilla V22025.1243.4
Dynamic DR (DTDR-C)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1243.3
No ICLBackbone=Gorilla V22025.1239.5
Learned RetrievalBackbone=Gorilla V22025.1237.9
DTDR-CBackbone=Qwen3 0.6B2025.1235.3
Dependency RetrievalBackbone=Gorilla V22025.1232.2
DR (Liu et al. 2024a)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1230.7
LR (Erdogan et al. 2024)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1225.6
QTS (Paramanayakam et al. 2025)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1223.7
No ICLBackbone=Qwen3 0.6B2025.1223.2
BM-25LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1223.1
Dependency RetrievalBackbone=Qwen3 0.6B2025.1222
QTS (Gao et al. 2025)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1221.5
Learned RetrievalBackbone=Qwen3 0.6B2025.1219.8
Raw DemonstrationsBackbone=Qwen3 0.6B2025.1217.2
QTS (Vanilla)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1215.8
Random GuessLLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.125.9