Spark 最近 24 小时 Digest

生成时间：2026-06-27 17:25 Asia/Shanghai

主题热度

  • agent: 19
  • engineering: 12
  • csdn: 11
  • systems: 11
  • other: 10
  • rag: 10
  • multimodal: 9
  • risk: 9
  • database: 5

可复用结论

  • [评审] Scaling Speech-Text Pre-training with Synthetic Interleaved Data：[评审] Scaling Speech-Text Pre-training with Synthetic Interleaved Data（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-scaling-speech-text-pre-training-with-synthetic-interleaved.md）
  • [评审] RaSA: Rank-Sharing Low-Rank Adaptation：[评审] RaSA: Rank-Sharing Low-Rank Adaptation（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-rasa-rank-sharing-low-rank-adaptation.md）
  • [评审] Optimal Transport for Time Series Imputation：[评审] Optimal Transport for Time Series Imputation（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-optimal-transport-for-time-series-imputation.md）
  • [评审] PIED: Physics-Informed Experimental Design for Inverse Problems：[评审] PIED: Physics-Informed Experimental Design for Inverse Problems（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-pied-physics-informed-experimental-design-for-inverse-proble.md）
  • [评审] DarkBench: Benchmarking Dark Patterns in Large Language Models：[评审] DarkBench: Benchmarking Dark Patterns in Large Language Models（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-darkbench-benchmarking-dark-patterns-in-large-language-model.md）
  • [评审] Common Corpus: The Largest Collection of Ethical Data for LLM Pre-Training：[评审] Common Corpus: The Largest Collection of Ethical Data for LLM Pre-Training（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-common-corpus-the-largest-collection-of-ethical-data-for-llm.md）
  • [评审] Decomposing LLM Computation with Jets：[评审] Decomposing LLM Computation with Jets（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-decomposing-llm-computation-with-jets.md）
  • [评审] Task Tokens: A Flexible Approach to Adapting Behavior Foundation Models：[评审] Task Tokens: A Flexible Approach to Adapting Behavior Foundation Models（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-task-tokens-a-flexible-approach-to-adapting-behavior-foundat.md）
  • [评审] THEMIS: Towards Holistic Evaluation of MLLMs for Scientific Paper Fraud Forensics：[评审] THEMIS: Towards Holistic Evaluation of MLLMs for Scientific Paper Fraud Forensics（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-themis-towards-holistic-evaluation-of-mllms-for-scientific-p.md）
  • [评审] RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style：[评审] RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style（来源：/shared/research-kb/inbox/flyp/2026-06-27-1650-rm-bench-benchmarking-reward-models-of-language-models-with.md）

建议进入主题页的要点

  • agent：知识库草稿：CSDN 高价值技术检索 · LLM API 工程 · Agent 记忆架构 · RAG 2026 新范式 · LoRA 微调排障 · 2026-06-27
  • rag：知识库草稿：CSDN 高价值技术检索 · LLM API 工程 · Agent 记忆架构 · RAG 2026 新范式 · LoRA 微调排障 · 2026-06-27
  • multimodal：Lilian Weng (Lil'Log) · RSS 摘要
  • systems：知识库草稿：CSDN 高价值技术检索 · LLM API 工程 · Agent 记忆架构 · RAG 2026 新范式 · LoRA 微调排障 · 2026-06-27
  • engineering：知识库草稿：CSDN 高价值技术检索 · LLM API 工程 · Agent 记忆架构 · RAG 2026 新范式 · LoRA 微调排障 · 2026-06-27
  • database：📋 简报 · 2026-06-27 下午 · Jay
  • csdn：知识库草稿：CSDN 高价值技术检索 · LLM API 工程 · Agent 记忆架构 · RAG 2026 新范式 · LoRA 微调排障 · 2026-06-27