模型思考时可改写记忆，但会拖慢学习速度。
Memoir: Should a Model Write to Its Memory While It Thinks?
- 让模型在推理时实时修改自身记忆内容
- 改写记忆导致学习速度下降约13.5%（显著更慢）
- 适合研究自洽推理与内存机制的学者
Memoir 结合了样本级快速记忆、共享慢参数、可变深度隐式循环和未来隐状态能量目标。我们测试其最激进的设计：每次思考迭代可能重写同一迭代读取的快速记忆层。在存在关键干扰的程序化关联回忆任务中，对比了耦合记忆（可改写）与仅读取记忆的对照组。两组均含81,738参数（其中76,362可训练），使用相同的乘加操作数、数据、优化器、调度和随机种子。经过240次训练步（12个种子），耦合记忆的准确率为0.5203（95%置信区间[0.4522, 0.5883]），而只读组为0.6557（[0.5953, 0.7160]）。配对t检验显示差异为0.1354，自由度11，95%置信区间[0.0431, 0.2277]，12个种子中有10个胜出。960步后两组均达1.0000，说明影响是学习速率延迟而非能力缺失。长期控制组已到上限，未测量非饱和任务下的收敛表现。预测的记忆重写破坏能量信号的情况并未发生：能量裕度持续增长并保持稳定。内核重构还将δ规则前向耗时从0.907毫秒降至0.351毫秒（指定设备）。代码与证据见https://github.com/RightNow-AI/Memoir。
原文摘要 · Abstract (English)
Memoir combines per-sample fast memory, shared slow parameters, variable-depth latent recurrence, and a future-latent energy objective. We test its riskiest coupling: each pondering iteration may rewrite the fast tier that the same iteration reads. On procedural associative recall with key interference, we compare a coupled arm against an otherwise identical read-only pondering arm. Both arms contain 81,738 parameters, including 76,362 trainable parameters, and use matched declared forward multiply-accumulate counts, data, optimizer, schedule, and seeds. After 240 training steps across 12 seeds, coupled recall is 0.5203 with a 95 percent interval of [0.4522, 0.5883], while read-only recall is 0.6557 with [0.5953, 0.7160]. The arms are paired per seed, and the read-only lead of 0.1354 gives a paired t of 3.23 on 11 degrees of freedom with a 95 percent interval of [0.0431, 0.2277] on the difference, winning on 10 of 12 seeds. After 960 steps across 8 seeds, both arms reach 1.0000, so the measured effect is a learning-speed penalty at a fixed budget, not a demonstrated capability penalty. That longer control is ceiling limited, leaving convergence on a non-saturating task unmeasured. A predicted failure in which memory rewriting corrupts the energy signal did not occur: the energy margin grew and held. Kernel restructuring also reduced delta-rule forward time from 0.907 ms to 0.351 ms on the stated device. Code and evidence are available at https://github.com/RightNow-AI/Memoir
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。