List your company

Function Selection on TaskBench HuggingFace

77.1Function Selection AccuracyRaw Demonstrations
1.28420.96740.6560.333Dec 18, 2025

Evaluation results

MethodLinks
Raw DemonstrationsBackbone=GPT-4o2025.1277.1
Raw DemonstrationsBackbone=Qwen3 14B2025.1276.7
No ICLBackbone=GPT-4o2025.1274.8
No ICLBackbone=Qwen3 14B2025.1274.4
Raw DemonstrationsBackbone=Qwen3 8B2025.1272.5
DTDR-CBackbone=GPT-4o2025.1271.9
DTDR-LBackbone=GPT-4o2025.1271.8
DTDR-CBackbone=Qwen3 14B2025.1271.5
DTDR-LBackbone=Qwen3 14B2025.1271.2
No ICLBackbone=Qwen3 8B2025.1270.5
Dependency RetrievalBackbone=GPT-4o2025.1270
Dependency RetrievalBackbone=Qwen3 14B2025.1269.2
Dependency RetrievalBackbone=Qwen3 8B2025.1267.3
DTDR-CBackbone=Qwen3 8B2025.1267.3
DTDR-LBackbone=Qwen3 8B2025.1263.8
DTDR-LBackbone=Qwen3 4B2025.1260.5
Learned RetrievalBackbone=Qwen3 4B2025.1256.9
DTDR-CBackbone=Qwen3 4B2025.1256
Learned RetrievalBackbone=GPT-4o2025.1255.2
Learned RetrievalBackbone=Qwen3 14B2025.1253.5
DTDR-CBackbone=Qwen3 1.7B2025.1252.8
Dependency RetrievalBackbone=Qwen3 4B2025.1252.8
Learned RetrievalBackbone=Qwen3 8B2025.1252
Raw DemonstrationsBackbone=Qwen3 4B2025.1249.8
DTDR-LBackbone=Qwen3 1.7B2025.1249
No ICLBackbone=Qwen3 4B2025.1245.8
Raw DemonstrationsBackbone=Qwen3 1.7B2025.1243.1
Learned RetrievalBackbone=Qwen3 1.7B2025.1242.8
Dependency RetrievalBackbone=Qwen3 1.7B2025.1241.6
No ICLBackbone=Qwen3 1.7B2025.1231.2
Dynamic LR (DTDR-L)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1227.8
Dynamic DR (DTDR-C)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1227.5
DTDR-CBackbone=Qwen3 0.6B2025.1220.8
LR (Erdogan et al. 2024)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1220.5
DTDR-LBackbone=Qwen3 0.6B2025.1218.9
DR (Liu et al. 2024a)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1217.1
QTS (Gao et al. 2025)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1214.6
QTS (Paramanayakam et al. 2025)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.1212.9
Learned RetrievalBackbone=Qwen3 0.6B2025.1210
BM-25LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.128.3
QTS (Vanilla)LLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.127.9
Dependency RetrievalBackbone=Qwen3 0.6B2025.126.7
No ICLBackbone=Qwen3 0.6B2025.125.5
Raw DemonstrationsBackbone=Qwen3 0.6B2025.124.3
Random GuessLLM Backbone=Qwen 3 0.6B, ICL Method=hard masking2025.124.2