ResearchBenchmarksRevision Generation on RESEARCHARCADEFollow0.733SBERT ScoreLLM0.304520.415760.5270.63824Nov 27, 2025Evaluation ResultsMethodMethodLinksSBERT ScoreROUGE-LGPT-4o-mini ScoreLLMBackbone=Qwen3-0.6B, T...Backbone=Qwen3-0.6B, Training=Trained, Hop=1-hop2025.110.7330.5540.572LLMBackbone=Qwen3-8B, Tra...Backbone=Qwen3-8B, Training=Fixed, Hop=1-hop2025.110.7040.4460.889LLMBackbone=GPTOSS-120B,...Backbone=GPTOSS-120B, Training=Fixed, Hop=1-hop2025.110.6690.2650.999LLMBackbone=Qwen3-0.6B, T...Backbone=Qwen3-0.6B, Training=Fixed, Hop=1-hop2025.110.3210.210.447