ResearchBenchmarksLong-context Question Answering on LongBench Code Repository Understanding v2For AI assistantsMarkdownContent indexLong-context Question Answering on LongBench Code Repository Understanding v2Follow52AccuracyRLM27.0433.524046.48Jul 1, 2026Evaluation resultsMethodMethodLinksAccuracyLatency (s)RLMBackbone=Qwen3.6-35B-A3BBackbone=Qwen3.6-35B-A3B2026.0752187.8MPLMBackbone=Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2026.073565.5RLMBackbone=Qwen3-30B-A3BBackbone=Qwen3-30B-A3B2026.0732.593.3MPLMBackbone=Qwen3.6-35B-A3BBackbone=Qwen3.6-35B-A3B2026.072856.7