List your company

Numerical Question Answering on FinLongDocQA 1.0 (test)

41.34EMFinLongDocAgent
-1.52889.600620.7331.8594Apr 4, 2026

Evaluation results

MethodLinks
FinLongDocAgentRetrieval Setting=Agentic RAG, Backbone LLM=Gemini-3-Flash, Retriever=BGE, Rounds=5, top-k=152026.0441.3443.5451.29
Gemini-3-FlashRetrieval Setting=With BGE retriever, Backbone LLM=Gemini-3-Flash2026.0433.0341.9349.9
FinLongDocAgentRetrieval Setting=Agentic RAG, Backbone LLM=DeepSeek-v3.2, Retriever=BGE, Rounds=5, top-k=152026.0431.0640.0347.3
FinLongDocAgentRetrieval Setting=Agentic RAG, Backbone LLM=Qwen3-Thinking 30B, Retriever=BGE, Rounds=5, top-k=152026.0430.137.1443.01
DeepSeek-v3.2Retrieval Setting=With BGE retriever, Backbone LLM=DeepSeek-v3.22026.0427.6638.7145.44
Qwen3-Thinking 30BRetrieval Setting=With BGE retriever, Backbone LLM=Qwen3-Thinking 30B2026.0427.2136.4742.43
WebDancer-30BRetrieval Setting=Agentic RAG, Retriever=BGE2026.0422.631.838.27
FinLongDocAgentRetrieval Setting=Agentic RAG, Backbone LLM=GPT-4o-mini, Retriever=BGE, Rounds=5, top-k=152026.0420.9730.5136.97
GPT-4o-miniRetrieval Setting=With BGE retriever, Backbone LLM=GPT-4o-mini2026.0419.0728.0936.4
Gemini-3-FlashRetrieval Setting=With report (long-context prompting), Backbone LLM=Gemini-3-Flash2026.0417.7921.9826.03
HippoRAG2Retrieval Setting=Graph-based RAG, Backbone LLM=GPT-4o-mini2026.0415.1222.7730.83
Qwen3-Thinking 30BRetrieval Setting=With report (long-context prompting), Backbone LLM=Qwen3-Thinking 30B2026.0413.8418.8823.16
DeepSeek-v3.2Retrieval Setting=With report (long-context prompting), Backbone LLM=DeepSeek-v3.22026.0413.4518.5122.25
GraphRAGRetrieval Setting=Graph-based RAG, Backbone LLM=GPT-4o-mini2026.0412.1418.1125.41
GPT-4o-miniRetrieval Setting=With report (long-context prompting), Backbone LLM=GPT-4o-mini2026.047.8811.7315.06
Gemini-3-FlashRetrieval Setting=Without context, Backbone LLM=Gemini-3-Flash2026.043.244.5218.06
DeepSeek-v3.2Retrieval Setting=Without context, Backbone LLM=DeepSeek-v3.22026.041.181.4710.39
Qwen3-Thinking 30BRetrieval Setting=Without context, Backbone LLM=Qwen3-Thinking 30B2026.040.240.284.09
GPT-4o-miniRetrieval Setting=Without context, Backbone LLM=GPT-4o-mini2026.040.120.130.56