List your company

Search-based Question Answering on SQA v2

87.7Overall ScoreAi2 ScholarQA
19.5837.26554.9572.635Jun 15, 2026

Evaluation results

MethodLinks
Ai2 ScholarQABackbone=Claude Sonnet2026.0687.788.189.192.481.2
DR Tulu-8B (1900-step, ≈9700h)Backbone=Qwen3-8B, Steps=1900-step, Compute Cost=≈9700h2026.0686.889.695.488.673.7
DR Tulu-8B (1000-step, ≈6000h)Backbone=Qwen3-8B, Steps=1000-step, Compute Cost=≈6000h2026.0686.784.895.490.676.1
DEEPRUBRIC (140-step, ≈750h)Backbone=Qwen3-8B, Steps=140-step, Compute Cost=≈750h2026.068686.49491.572.2
DEEPRUBRIC (75-step, ≈402h)Backbone=Qwen3-8B, Steps=75-step, Compute Cost=≈402h2026.0685.185.393.789.372.1
OpenAI Deep ResearchBackbone=o3/o42026.0679.691.595.677.443.1
DEEPRUBRIC (SFT)Backbone=Qwen3-8B, Stage=SFT2026.0679.578.494.283.961.4
GPT-5 + SearchBackbone=GPT-52026.0674.892.393.867.845.6
DR Tulu-8B (SFT)Backbone=Qwen3-8B2026.0672.381.49165.351.6
Gemini 3 Pro + SearchBackbone=Gemini 3 Pro2026.0669.883.198.368.529.4
Perplexity Deep ResearchBackbone=Proprietary2026.0667.391.692.747.337.6
Qwen3-8B + SearchBackbone=Qwen3-8B2026.0657.242.892.153.740.3
WebThinker-32B-DPO (report)Backbone=Qwen2.5-32B (report)2026.0646.791.295.5——
Tongyi DeepResearch-30B-A3BBackbone=Qwen2.5-30B2026.0646.589.596.4——
WebThinker (report)Backbone=QwQ-32B (report)2026.0645.286.494.3——
WebExplorer-8BBackbone=Qwen2.5-8B2026.0642.578.691.4——
QwQ-32B + RAGBackbone=QwQ-32B2026.0641.977.590.3——
Qwen3-8B + RAGBackbone=Qwen3-8B2026.0640.469.292.3——
WebThinker-32B-DPOBackbone=Qwen2.5-32B2026.0632.936.794.9——
ASearcher-Web-7BBackbone=Qwen2.5-7B2026.0626.913.794——
Search-R1-7BBackbone=Qwen2.5-7B2026.0622.29.779——