16 个记忆 token 让冻结 LLM 超越全上下文推理
论文把整套文档压成 16 个 token 喂给冻结 LLM，FactKG 上反超全上下文，还能当智能体记忆用，思路很巧。
一篇论文提出 Latent Semantic Memory 方法，把文档编码为可复用的嵌入，再由学习模块将相关信息压缩写入 16 个 token 输入冻结 LLM。在 FactKG 基准上得分 84.33%，超过全上下文推理的 80.43%，源文档 token 用量减少 108 倍。在八个基准中的五个上击败 Mem0 和 MemoRAG 等基线。作为智能体记忆使用时，TextWorld 中成功率从 18.6% 提升到 24.2%。
// Latent Semantic Memory for LLMs // This paper presents some really interesting findings about agent memory. They find that sixteen memory tokens beat full-context inference on FactKG while using 108x fewer tokens for the source documents. Approach: Documents are encoded once into reusable embeddings. For each query, a learned module gathers the relevant information from the whole collection and writes it into 16 tokens in the input space of a frozen LLM. Results: It beats Mem0, MemoRAG, and other baselines on five of eight benchmarks. On FactKG, it scores 84.33% against 80.43% with the full context. It also works as agent memory. In TextWorld, using the agent's own history as the collection raises success from 18.6% to 24.2%, and the agent revisits fewer rooms. It also seems that it could help frozen LLMs accumulate and use knowledge effectively across long-context interactions. This is applicable to building continually learning AI systems. arxiv.org/abs/2610.21375 Chat with Paper: academy.dair.ai/papers/memoril… 💬 9 🔄 3 ❤️ 14 👀 1649 📊 11 ⚡
