Event ·Multi-source

Context–Answer Metamodels:在首个答案 token 生成前预测 LLM 激活

一篇 arXiv 论文提出 context–answer metamodels,把模型读上下文时的激活序列映射为答案激活的点预测或分布预测。实验覆盖 4 种 metamodel 家族,训练集规模扩展到 500,000 条 prompts 时预测精度持续提升,且能零样本泛化到域外数据集。预测结果保留了行为层面的预测信号。Qwen2.5 的上下文激活可以预测 Qwen3.5 和 Gemma 4 的答案激活,说明激活预测能跨模型迁移。

Current read

用读上下文的激活就能预测模型答案阶段的激活,跨 Qwen2.5 到 Gemma 4 都能迁移,做可解释性或行为预测的可以看看这篇。

5 sources31° AI heatLast updated 2026/11/4 17:53:52

Evidence

5 sources
Related source 2Latent.Space
View source
Related source 4rohanpaul_ai
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Context–Answer Metamodels:在首个答案 token 生成前预测 LLM 激活 · 事件专题 | AITOP