ResearchBenchmarksScience Question Answering on AI2ARC (test)Follow96.3AccuracyGPT-480.273684.434388.59592.7557Mar 15, 2023Jul 6, 2023Oct 28, 2023Feb 18, 2024Jun 11, 2024Oct 2, 2024Jan 24, 2025Evaluation ResultsMethodMethodLinksAccuracyGPT-4number_of_shots=25-shotnumber_of_shots=25-shot2023.0396.3ST-MoEbenchmark-specific tun...benchmark-specific tuning=true2023.0386.5GPT-3.5number_of_shots=25-shotnumber_of_shots=25-shot2023.0385.2PaLMnumber_of_shots=8-shotnumber_of_shots=8-shot2023.0385.2TAMA2025.0181.23base2025.0180.89