Locked at the Entrance, Open Inside: Where RLVR Narrows the Solution Space

  • 类型:arxiv
  • 标识:2608.29188
  • 链接:https://arxiv.org/abs/2608.29188
  • 主分类:llm-infra
  • 形态:method
  • 被引:0
  • 被引来源:Semantic Scholar + OpenAlex
  • S2被引:0
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:While surface prompting fails to recover diversity, entrance-targeted interventions succeed: late-layer parameter interpolation with early checkpoints increases solution coverage by 37% at no loss in pass@1 and late-layer parameter interpolation with early checkpoints increases solution coverage by 37% at no loss in pass@1.
  • OpenAlex ID:W7204921312
  • OpenAlex DOI:10.48550/arxiv.2608.29188
  • DOI:10.48550/arxiv.2608.29188
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/2608.29188
  • OpenAlex更新:2026-09-07
  • 待LLM分类:否
  • 标题中文:锁于入口,开于内里:RLVR 收窄解空间的位置
  • TLDR中文:表面 prompting 未能恢复多样性,而针对入口的干预则成功:使用早期 checkpoint 的 late-layer parameter interpolation 在不损失 pass@1 的情况下将解的覆盖度提高了 37%。
  • 成熟度:research
  • 场景:RLVR、reasoning、test-time scaling
  • 来源文件:
  • /inbox/tom/_candidates/2026-09-05-agent-rag-longcontext-candidates.json
  • /inbox/tom/_candidates/2026-09-06-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]
  • /inbox/tom/_candidates/2026-09-07-agent-rag-longcontext-candidates.json