ResearchBenchmarksGraduate-Level Question Answering on GPQA (Acceptance length, Delta (%))Follow2Acceptance LengthTTS1.1681.3841.61.816May 10, 2026Evaluation ResultsMethodMethodLinksAcceptance LengthDelta (%)TTSModel=Llama3.1-8BModel=Llama3.1-8B2026.05260.4TTSModel=Qwen/Qwen3-8BModel=Qwen/Qwen3-8B2026.051.924.5TTSModel=Qwen/Qwen3-32BModel=Qwen/Qwen3-32B2026.051.821.5EAGLE-3Model=Qwen/Qwen3-8BModel=Qwen/Qwen3-8B2026.051.5—EAGLE-3Model=Qwen/Qwen3-32BModel=Qwen/Qwen3-32B2026.051.4—EAGLE-3Model=Llama3.1-8BModel=Llama3.1-8B2026.051.2—