按年被提到
20252026 · 2 次2026
大模型与语言模型›量化、解码与分词·arXiv:2511.05963 · cs.LG

Next-Latent Prediction Transformers Learn Compact World Models

Jayden Teoh, Manan Tomar, Kwangjun Ahn, Edward S. Hu, Tim Pearce, Pratyusha Sharma, Akshay Krishnamurthy, Riashat Islam, Alex Lamb, John Langford · 2025
开始阅读 ↗PDFalphaXiv 讨论
被提及 · 2 人首次 2026-06-17 · 最近 2026-10-12
2不同的人
2活跃月份
2讨论串
4被引(OpenAlex)
arXiv 摘要(候选层不写导读)

Transformers replace recurrence with a memory that grows with sequence length and self-attention that enables ad-hoc lookups over past tokens. Consequently, they lack an inherent incentive to compress history into compact latent states with consistent transition rules. This often leads to learning solutions that generalize poorly. We introduce Next-Latent Prediction (NextLat), which extends standard next-token training with self-supervised predictions in the latent space. Specifically, NextLat trains a transformer to learn latent representations that are predictive of its next latent state given the next token. Theoretically, we show that these latents provably converge towards belief states, compressed information about the history necessary to predict the future. This simple auxiliary objective injects a recurrent inductive bias into transformers while leaving their architecture, parallel training efficiency, and inference unchanged. NextLat effectively encourages transformers to form compact internal world models with coherent belief states and transition dynamics -- crucial properties not guaranteed by standard next-token prediction alone. Empirically, across benchmarks in world modeling, reasoning, planning, and language modeling, NextLat demonstrates significant gains over standard next-token prediction and other baselines in downstream accuracy, representation compression, and lookahead planning. Furthermore, NextLat enables variable-length self-speculative decoding, accelerating inference by up to 3.3x in language modeling. NextLat offers a simple yet effective paradigm for learning compact, predictive representations in transformers that generalize better. Our code is available at https://github.com/JaydenTeoh/NextLat.

候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
评论jlhawn · 2026-10

“> LLMs are powerful autocomplete with vast and broad memory, they do not think. This is really only true insofar as they are designed such that they output language tokens and their intermediate "chain of thought" is also done in language tokens, but now that's really only done for auditability. There are much more abstract layers before the model output…”

Thinking will become a hobby原评论 ↗
被提交为 story · 1 次