Information Retrieval on BRIGHT
54.8Mean nDCG@10qwen3-4b
Evaluation Results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| qwen3-4bEnhancement=CoT Reasoner2026.05 | 54.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| qwen3-4bEnhancement=Reasoning Internalizer2026.05 | 51.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| qwen3-4bEnhancement=None2026.05 | 51.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| snowflakeEnhancement=CoT Reasoner2026.05 | 44 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| gte-baseEnhancement=CoT Reasoner2026.05 | 43.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| e5-largeEnhancement=CoT Reasoner2026.05 | 43.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| gte-baseEnhancement=Reasoning Internalizer2026.05 | 39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| snowflakeEnhancement=Reasoning Internalizer2026.05 | 38.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReasonEmbed-Qwen3-8BSize=8B, Loss function=Redapter (using the self-adaptive RI-InfoNCE loss)2025.10 | 38.1 | 55.5 | 56.6 | 36.2 | 47.4 | 35.3 | 36.6 | 39.1 | 16.4 | 33.6 | 12.5 | 47.2 | 41.4 | — | — | — | — | |
| e5-largeEnhancement=Reasoning Internalizer2026.05 | 37.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReasonEmbed-Qwen3-8BSize=8B, Loss function=basic contrastive learning (using InfoNCE loss)2025.10 | 37.1 | 54.4 | 55.4 | 33.8 | 45.2 | 32 | 34.3 | 37.3 | 18.7 | 32.3 | 13.3 | 47.6 | 41.2 | — | — | — | — | |
| ReasonEmbed-Qwen3-4BSize=4B, Loss function=Redapter (using the self-adaptive RI-InfoNCE loss)2025.10 | 37.1 | 55.4 | 54.5 | 34.9 | 46.9 | 34 | 36.1 | 37.4 | 13.6 | 34.5 | 11.3 | 45.1 | 41.4 | — | — | — | — | |
| gte-baseEnhancement=None2026.05 | 37 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReasonEmbed-Llama-3.1-8BSize=8B, Loss function=Redapter (using the self-adaptive RI-InfoNCE loss)2025.10 | 36.2 | 55.4 | 56.2 | 35.2 | 48.5 | 32.1 | 37.3 | 41.1 | 16.8 | 28.8 | 9.1 | 36.6 | 37.9 | — | — | — | — | |
| ReasonEmbed-Qwen3-4BSize=4B, Loss function=basic contrastive learning (using InfoNCE loss)2025.10 | 35.3 | 51.8 | 53.3 | 34.1 | 42.6 | 31.1 | 32.1 | 35.6 | 11.6 | 32.7 | 13 | 45.2 | 40.8 | — | — | — | — | |
| ReasonEmbed-Llama-3.1-8BSize=8B, Loss function=basic contrastive learning (using InfoNCE loss)2025.10 | 34.9 | 55.5 | 53.8 | 36.4 | 45.6 | 29.8 | 35.1 | 38.6 | 12.5 | 29.4 | 9.5 | 35.7 | 36.7 | — | — | — | — | |
| snowflakeEnhancement=None2026.05 | 34.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| e5-largeEnhancement=None2026.05 | 31.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LaSERBackbone=Qwen3-8B, Size=8B2026.03 | 29.3 | 58 | 51.8 | 29 | 44.1 | 26 | 32.9 | 34.5 | 11.7 | 9.2 | 1.6 | 34 | 18.7 | — | — | — | — | |
| DIVER-RetrieverSize=4B, Method Type=Tailored methods for reasoning-intensive retrieval2025.10 | 28.9 | 41.8 | 43.7 | 21.7 | 35.3 | 21 | 21.2 | 25.1 | 13.2 | 37.6 | 10.7 | 37.3 | 38.4 | — | — | — | — | |
| Rewrite-then-RetrieveBackbone=Qwen3-8B, Size=8B, external_LLM_query_rewriting=true2026.03 | 28.1 | 53.1 | 54.3 | 32.1 | 34.8 | 20.5 | 31.1 | 32.2 | 15.2 | 3.2 | 4.1 | 38.8 | 17.4 | — | — | — | — | |
| Seed-1.5-EmbeddingMethod Type=Tailored methods for reasoning-intensive retrieval2025.10 | 27.2 | 34.8 | 46.9 | 23.4 | 31.6 | 19.1 | 25.4 | 21 | 4.9 | 43.2 | 12.2 | 30.5 | 33.3 | — | — | — | — | |
| GRC-4BSize=4B, Temperature=0.6, Max Tokens=4K2026.05 | 27.2 | 33.5 | 32 | 26.4 | 30.6 | 19.6 | 28.2 | 24.8 | 4.1 | 38.3 | 11.3 | 33.9 | 44.2 | — | — | — | — | |
| Rewrite-then-RetrieveBackbone=LLaMA3.1-8B, Size=8B, external_LLM_query_rewriting=true2026.03 | 26.9 | 63.1 | 54.5 | 27.7 | 40.6 | 17.1 | 28.1 | 28.2 | 11.4 | 1.6 | 3.8 | 30.3 | 16.1 | — | — | — | — | |
| BracketRank-20Model=GPT-42026.04 | 26.6 | 38.2 | 45.2 | 22.7 | 36.5 | 30.5 | 26.2 | 33.8 | 28.7 | 16.9 | 9.2 | 16.6 | 14.2 | — | — | — | — | |
| GIRCSEBackbone=Qwen3-8B, Size=8B2026.03 | 26 | 59 | 56.5 | 27.2 | 40.3 | 19 | 28.5 | 31.4 | 3.6 | 3.2 | 1.7 | 27.2 | 14 | — | — | — | — | |
| Fair BaselineBackbone=Qwen3-8B, Size=8B2026.03 | 25.7 | 49.7 | 51.2 | 26.9 | 37.4 | 23.4 | 28 | 34.1 | 3.2 | 3.7 | 2.8 | 31.8 | 16.8 | — | — | — | — | |
| GRC-1.7BSize=1.7B, Temperature=0.6, Max Tokens=4K2026.05 | 25.7 | 37.5 | 37.6 | 24.5 | 31.3 | 18.9 | 23.1 | 25.7 | 11.2 | 26.2 | 9 | 26 | 37.3 | — | — | — | — | |
| RaDeR-gte-Qwen2-7BSize=7B, Method Type=Tailored methods for reasoning-intensive retrieval2025.10 | 25.5 | 34.6 | 38.9 | 22.1 | 33 | 14.8 | 22.5 | 23.7 | 5 | 37.3 | 10.2 | 35.1 | 28.4 | — | — | — | — | |
| LaSERBackbone=LLaMA3.1-8B, Size=8B2026.03 | 24.4 | 58.4 | 48.1 | 28 | 40.9 | 17 | 29.9 | 28.3 | 5.9 | 1.7 | 1.5 | 19.2 | 14.6 | — | — | — | — | |
| ReasonIR-8BSize=8B, Method Type=Tailored methods for reasoning-intensive retrieval2025.10 | 24.4 | 26.2 | 31.4 | 23.3 | 30 | 18 | 23.9 | 20.5 | 10.5 | 35 | 14.7 | 27.2 | 31.9 | — | — | — | — | |
| ReasonIR-8BSize=8B2026.05 | 24.4 | 26.2 | 31.4 | 23.3 | 30.3 | 18 | 23.9 | 20.5 | 10.5 | 35 | 14.7 | 27.2 | 31.9 | — | — | — | — | |
| gte-Qwen2-7B-instructSize=7B, Method Type=General-purpose methods2025.10 | 23.5 | 34.1 | 42.6 | 18.2 | 27.4 | 13.2 | 17.3 | 20.9 | 2.2 | 30.4 | 13.3 | 32.6 | 30.6 | — | — | — | — | |
| LaSERBackbone=Qwen3-0.6B, Size=0.6B2026.03 | 23.1 | 50 | 45.9 | 25.7 | 32.4 | 18.4 | 27.1 | 26.5 | 2.7 | 9.1 | 1.2 | 22.4 | 16 | — | — | — | — | |
| Qwen3-Embedding-8B*Size=8B, Method Type=General-purpose methods2025.10 | 22.8 | 21 | 33 | 18.4 | 26.1 | 15.7 | 19.4 | 17.3 | 1.2 | 33.8 | 9.4 | 39.3 | 39.2 | — | — | — | — | |
| Fair BaselineBackbone=LLaMA3.1-8B, Size=8B2026.03 | 22.5 | 57.7 | 40.9 | 24.6 | 29.8 | 20.1 | 29.2 | 25.8 | 5.7 | 4.3 | 1.6 | 17.7 | 12 | — | — | — | — | |
| QwenSize=7.7B2026.05 | 22.5 | 30.6 | 36.4 | 17.8 | 24.6 | 13.2 | 22.2 | 14.8 | 9.9 | 25.5 | 14.4 | 32.9 | 27.8 | — | — | — | — | |
| Rewrite-then-RetrieveBackbone=Qwen3-0.6B, Size=0.6B, external_LLM_query_rewriting=true2026.03 | 22.4 | 57.8 | 51.6 | 14.1 | 39.4 | 15 | 19.8 | 23.7 | 14.6 | 0.6 | 1.2 | 15.5 | 14.9 | — | — | — | — | |
| GIRCSEBackbone=LLaMA3.1-8B, Size=8B2026.03 | 22 | 50.6 | 43.4 | 28.3 | 35.7 | 14.3 | 26.9 | 26.4 | 5.2 | 6 | 0.5 | 15.1 | 11.6 | — | — | — | — | |
| Qwen3-Embedding-4B*Size=4B, Method Type=General-purpose methods2025.10 | 21.8 | 17.8 | 34.7 | 16.9 | 23.3 | 12.5 | 16.2 | 16.8 | 1.4 | 35.7 | 9.8 | 41.5 | 35.5 | — | — | — | — | |
| GritLM-7BSize=7B, Method Type=General-purpose methods2025.10 | 21 | 24.8 | 32.3 | 18.9 | 19.8 | 17.1 | 13.6 | 17.8 | 22 | 29.9 | 8.8 | 21.2 | 25.2 | — | — | — | — | |
| GritLM-7BSize=7.1B2026.05 | 21 | 24.8 | 32.3 | 18.9 | 19.8 | 17.1 | 13.6 | 17.8 | 22 | 29.9 | 8.8 | 21.2 | 25.2 | — | — | — | — | |
| Rank-R1-14BProtocol=Distilled / Fine-tuned, Training=GRPO2026.04 | 20.5 | 31.2 | 38.5 | 21.2 | 26.4 | 22.6 | 18.9 | 27.5 | 9.2 | 20.2 | 9.7 | 11.9 | 9.2 | — | — | — | — | |
| RevelaModel Size=3B2025.06 | 20.1 | 24.9 | 40.3 | 17.2 | 21.4 | 15.3 | 16.9 | 19.2 | 6.5 | 26.6 | 11.6 | 14.2 | 27.1 | — | — | — | — | |
| Google-Gecko-1B-768Size=1B, Method Type=General-purpose methods2025.10 | 20 | 22.7 | 34.8 | 19.6 | 27.8 | 15.7 | 20.1 | 17.1 | 3.6 | 29.6 | 9.3 | 15.9 | 23.8 | — | — | — | — | |
| GoogleSize=N/A2026.05 | 20 | 22.7 | 34.8 | 19.6 | 27.8 | 15.7 | 20.1 | 17.1 | 3.6 | 29.6 | 9.3 | 15.9 | 23.8 | — | — | — | — | |
| Qwen3-4B-ms†Size=4B, Method Type=General-purpose methods2025.10 | 19.4 | 16.8 | 34.9 | 16 | 19.4 | 20.1 | 18.8 | 11.1 | 9.1 | 33.3 | 8.7 | 18 | 27 | — | — | — | — | |
| Rank-R1-14BProtocol=Zero-Shot2026.04 | 19.3 | 30.1 | 36.6 | 22.1 | 24.6 | 21.7 | 15.4 | 25 | 9 | 17 | 9.1 | 11.6 | 9.2 | — | — | — | — | |
| GIRCSEBackbone=Qwen3-0.6B, Size=0.6B2026.03 | 19.1 | 29 | 32.8 | 24.7 | 30.6 | 13.6 | 24 | 26.6 | 1.1 | 11.1 | 1.3 | 20.6 | 13.5 | — | — | — | — | |
| Inst-XLSize=1.5B2026.05 | 18.9 | 21.6 | 34.3 | 22.4 | 27.4 | 18.2 | 21.2 | 19.1 | 5 | 27.5 | 8.5 | 5.9 | 15.6 | — | — | — | — | |
| Qwen3-8B-ms†Size=8B, Method Type=General-purpose methods2025.10 | 18.7 | 15.9 | 35.6 | 16.1 | 19.1 | 20.5 | 19.2 | 11.7 | 9.5 | 29.2 | 8.4 | 17 | 22.2 | — | — | — | — | |
| Fair BaselineBackbone=Qwen3-0.6B, Size=0.6B2026.03 | 18.3 | 28.8 | 31.6 | 25.2 | 28.1 | 15.1 | 22.3 | 24.2 | 3.5 | 8.8 | 2.1 | 15.4 | 14.1 | — | — | — | — | |
| SFRSize=7.1B2026.05 | 18.3 | 19.1 | 26.7 | 17.8 | 19 | 16.3 | 14.4 | 19.2 | 2 | 27.4 | 7.4 | 26 | 24.3 | — | — | — | — | |
| Setwise-14BProtocol=Zero-Shot2026.04 | 18 | 29.5 | 32.2 | 20.5 | 24.8 | 18.9 | 14.7 | 23.6 | 8.7 | 18.7 | 8 | 9.3 | 7.6 | — | — | — | — | |
| E5-MistralModel Size=7B2025.06 | 17.9 | 18.6 | 26 | 15.5 | 15.8 | 16.3 | 11.2 | 18.1 | 4.9 | 28.7 | 7.1 | 26.8 | 26.1 | — | — | — | — | |
| OpenAIModel Size=–2025.06 | 17.9 | 23.3 | 26.7 | 19.5 | 27.6 | 12.8 | 14.3 | 20.5 | 2.4 | 23.6 | 8.5 | 11.7 | 23.5 | — | — | — | — | |
| VoyageModel Size=–2025.06 | 17.9 | 23.1 | 25.4 | 19.9 | 24.9 | 10.8 | 16.8 | 15.4 | 1.5 | 30.6 | 7.5 | 11.6 | 27.4 | — | — | — | — | |
| Qwen3-EmbeddingBackbone=Qwen3-Embedding-4B, Size=4B2026.03 | 17.9 | 15.9 | 33.6 | 16.7 | 23.2 | 13.2 | 15 | 16.8 | 1.7 | 21.2 | 4 | 35.4 | 18.4 | — | — | — | — | |
| OpenAI-3-LargeMethod Type=General-purpose methods2025.10 | 17.9 | 23.3 | 26.7 | 19.5 | 27.6 | 12.8 | 14.3 | 20.5 | 2.4 | 23.6 | 8.5 | 11.7 | 23.5 | — | — | — | — | |
| OpenAISize=N/A2026.05 | 17.9 | 23.3 | 26.7 | 19.5 | 27.6 | 12.8 | 14.3 | 20.5 | 2.4 | 23.6 | 8.5 | 11.7 | 23.5 | — | — | — | — | |
| VoyageSize=N/A2026.05 | 17.9 | 23.1 | 25.4 | 19.9 | 24.9 | 10.8 | 16.8 | 15.4 | 1.5 | 30.6 | 7.5 | 11.6 | 27.4 | — | — | — | — | |
| E5Size=7.1B2026.05 | 17.9 | 18.6 | 26 | 15.5 | 15.8 | 16.3 | 11.2 | 18.1 | 4.9 | 28.7 | 7.1 | 26.8 | 26.1 | — | — | — | — | |
| RankGPT-4†Protocol=Zero-Shot2026.04 | 17 | 33.8 | 34.2 | 16.7 | 27 | 22.3 | 27.7 | 11.1 | 15.6 | 3.4 | 1.2 | 8.6 | 0.2 | — | — | — | — | |
| Setwise-7BProtocol=Distilled / Fine-tuned, Training=SFT2026.04 | 16.7 | 28.7 | 30.1 | 14.1 | 23.9 | 18.9 | 13.7 | 19.6 | 7.1 | 20.7 | 7 | 8.2 | 8.2 | — | — | — | — | |
| Setwise-14BProtocol=Distilled / Fine-tuned, Training=SFT2026.04 | 16.7 | 22 | 29.3 | 15.4 | 23 | 20.1 | 15.7 | 20.3 | 6.2 | 19.4 | 9.5 | 9.9 | 9.7 | — | — | — | — | |
| CohereModel Size=–2025.06 | 16.6 | 18.7 | 28.4 | 20.4 | 21.6 | 16.3 | 18.3 | 17.6 | 1.9 | 26.8 | 6.3 | 7.2 | 15.7 | — | — | — | — | |
| CohereSize=N/A2026.05 | 16.6 | 18.7 | 28.4 | 20.4 | 21.6 | 16.3 | 18.3 | 17.6 | 1.9 | 26.8 | 6.3 | 7.2 | 15.7 | — | — | — | — | |
| E5-Mistral-7B-InstructSize=7B2026.03 | 16.5 | 17.9 | 28.6 | 18.5 | 16.5 | 18.6 | 13.4 | 20 | 6.3 | 19.7 | 2 | 22.1 | 14.2 | — | — | — | — | |
| Rank-R1-7BProtocol=Distilled / Fine-tuned, Training=GRPO2026.04 | 16.4 | 26 | 28.5 | 17.2 | 24.2 | 19.1 | 10.4 | 24.2 | 4.3 | 19.8 | 4.3 | 10.9 | 8.3 | — | — | — | — | |
| Llama-3.1-8B-ms†Size=8B, Method Type=General-purpose methods2025.10 | 16.1 | 12.2 | 29.3 | 14.3 | 16.7 | 14.7 | 14.4 | 11.6 | 3.5 | 28.9 | 8.7 | 14.9 | 23.7 | — | — | — | — | |
| RevelaModel Size=1B2025.06 | 15.6 | 16.7 | 29.4 | 13.2 | 15 | 10.5 | 15.5 | 15.4 | 3 | 26.1 | 8.8 | 8.9 | 24.1 | — | — | — | — | |
| RevelaModel Size=0.5B2025.06 | 15.2 | 16.9 | 29 | 13.9 | 13.3 | 9.1 | 10.8 | 13.9 | 3.9 | 27.6 | 12.5 | 6.7 | 24.8 | — | — | — | — | |
| Rank-R1-7BProtocol=Zero-Shot2026.04 | 15 | 26.8 | 24.8 | 17.9 | 22.1 | 17.4 | 10.3 | 21.1 | 4.4 | 15.6 | 3.3 | 10.4 | 5.9 | — | — | — | — | |
| BM25Qcategory=New2025.09 | 14.8 | — | — | — | — | — | — | — | — | — | — | — | — | 39 | — | — | — | |
| NAFcategory=New2025.09 | 14.8 | — | — | — | — | — | — | — | — | — | — | — | — | 38.1 | — | — | — | |
| RRFcategory=New2025.09 | 14.7 | — | — | — | — | — | — | — | — | — | — | — | — | 38.4 | — | — | — | |
| BM25Model Size=–2025.06 | 14.5 | 18.9 | 27.2 | 14.9 | 12.5 | 13.6 | 18.4 | 15 | 7.9 | 24.4 | 6.2 | 4.9 | 10.4 | — | — | — | — | |
| BM25Method Type=General-purpose methods2025.10 | 14.5 | 18.9 | 27.2 | 14.9 | 12.5 | 13.6 | 18.4 | 15 | 7.9 | 24.4 | 6.2 | 4.9 | 10.4 | — | — | — | — | |
| BM25Size=N/A2026.05 | 14.5 | 18.9 | 27.2 | 14.9 | 12.5 | 13.6 | 18.4 | 15 | 7.9 | 24.4 | 6.2 | 4.9 | 10.4 | — | — | — | — | |
| Qwen3-EmbeddingBackbone=Qwen3-Embedding-0.6B, Size=0.6B2026.03 | 14.4 | 12.7 | 26.3 | 17.9 | 16.5 | 12.5 | 12.4 | 12.2 | 0.7 | 14.3 | 3.1 | 26.5 | 17.2 | — | — | — | — | |
| Qwen3-EmbeddingBackbone=Qwen3-Embedding-8B, Size=8B2026.03 | 14 | 14.7 | 17.9 | 15.5 | 19.9 | 9.1 | 12.9 | 16.5 | 0.8 | 17.4 | 2.5 | 24.5 | 16.8 | — | — | — | — | |
| BM252026.04 | 13.7 | 18.2 | 27.9 | 16.4 | 13.4 | 10.9 | 16.3 | 16.1 | 4.3 | 24.7 | 6.5 | 2.1 | 7.3 | — | — | — | — | |
| BoWcategory=Baseline2025.09 | 13.7 | — | — | — | — | — | — | — | — | — | — | — | — | 34.7 | — | — | — | |
| Setwise-7BProtocol=Zero-Shot2026.04 | 13.6 | 23.6 | 22.3 | 16.1 | 17.1 | 14.9 | 9.2 | 18.3 | 6.3 | 14.9 | 4.1 | 10.4 | 5.6 | — | — | — | — | |
| E5-PTModel Size=0.3B2025.06 | 13.1 | 18.5 | 29.5 | 10.2 | 17.6 | 7.1 | 8.7 | 14.2 | 1.8 | 22.5 | 3.3 | 6 | 17.4 | — | — | — | — | |
| RankZephyr-7BProtocol=Distilled / Fine-tuned2026.04 | 13 | 21.9 | 23.7 | 14.4 | 10.3 | 7.6 | 13.7 | 16.6 | 6.5 | 24.7 | 6.8 | 2 | 7.3 | — | — | — | — | |
| Multilingual-E5-Large-InstructSize=0.6B2026.03 | 12.3 | 15 | 24.6 | 14 | 14.6 | 16.1 | 10.3 | 13.8 | 2.1 | 15.1 | 3 | 6.9 | 11.6 | — | — | — | — | |
| REPLUGModel Size=3B2025.06 | 12.2 | 6.8 | 13 | 12.9 | 15.1 | 7.4 | 7.5 | 12.1 | 1.1 | 24.8 | 13.3 | 9.2 | 22.9 | — | — | — | — | |
| BGE-M3Size=0.6B2026.03 | 11.4 | 7.2 | 13.3 | 12.5 | 12.8 | 12.6 | 10 | 10.1 | 30.7 | 15.6 | 1.5 | 5 | 5.6 | — | — | — | — | |
| Rank-R1-3BProtocol=Distilled / Fine-tuned, Training=GRPO2026.04 | 10.8 | 18.4 | 17.1 | 13.7 | 16.9 | 9 | 10 | 16.5 | 4.7 | 11.1 | 3.5 | 5.9 | 3.2 | — | — | — | — | |
| Setwise-3BProtocol=Distilled / Fine-tuned, Training=SFT2026.04 | 10.1 | 22 | 18.8 | 10.4 | 11.5 | 9.1 | 5.8 | 16.7 | 5.7 | 9.9 | 4 | 3.4 | 3.8 | — | — | — | — | |
| Rank-R1-3BProtocol=Zero-Shot2026.04 | 9.7 | 13.7 | 17.3 | 11.9 | 15.2 | 10 | 6.6 | 17.8 | 3.7 | 7.7 | 4 | 6 | 2.5 | — | — | — | — | |
| Setwise-3BProtocol=Zero-Shot2026.04 | 9.5 | 14.3 | 17.5 | 12 | 10.2 | 7.7 | 7.9 | 15.4 | 5.3 | 15.4 | 1.7 | 4.2 | 2.1 | — | — | — | — | |
| Search-R3Backbone=Qwen2.5-1.5B, Size=1.5B2026.03 | 7.7 | 13.8 | 8.3 | 4.2 | 3.5 | 4.5 | 12.4 | 4.6 | 3 | 16.7 | 0.6 | 11.7 | 8.5 | — | — | — | — | |
| BM252026.05 | — | 18.9 | 27.2 | 14.9 | — | 13.6 | — | — | — | — | — | — | — | — | — | — | — | |
| DCI-Agent-CCBackbone=Sonnet 4.62026.05 | — | 77.1 | 69 | 46.8 | — | 56.8 | — | — | — | — | — | — | — | — | — | — | — | |
| DCI-Agent-LiteBackbone=GPT-5.4 nano2026.05 | — | 60 | 50.8 | 32.3 | — | 42.4 | — | — | — | — | — | — | — | — | — | — | — | |
| GTE-Qwen2-7B-InstructBackbone=Qwen2, Model Size=7B2026.05 | — | 30.6 | 36.4 | 17.8 | — | 13.2 | — | — | — | — | — | — | — | — | — | — | — | |
| OpenAI-text-emb-3-large2026.05 | — | 23.3 | 26.7 | 19.5 | — | 12.8 | — | — | — | — | — | — | — | — | — | — | — | |
| Rank-R1-14BModel Size=14B2026.05 | — | 31.2 | 38.5 | 21.2 | — | 22.6 | — | — | — | — | — | — | — | — | — | — | — |