arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.14751cs.LGcs.AI
收藏

超越多令牌预测:通过未来摘要预训练语言模型

Beyond Multi-Token Prediction: Pretraining LLMs with Future Summaries

Carnegie Mellon University(卡内基梅隆大学);FAIR at Meta(Meta的FAIR)

Divyat Mahajan, Sachin Goyal, Badr Youbi Idrissi, Mohammad Pezeshki, Ioannis Mitliagkas, David Lopez-Paz, Kartik Ahuja

AI总结:

本文提出未来摘要预测(FSP)方法,通过训练辅助头预测紧凑的长期未来表示,提升长形式生成能力,在数学、推理和编码基准测试中优于NTP和MTP。

摘要:

Next-token prediction (NTP) 驱动了大型语言模型 (LLMs) 的成功,但其在长距离推理、规划和创造性写作中存在局限,主要归因于教师强制训练。多令牌预测 (MTP) 部分缓解了这些问题,但主要捕捉短程依赖关系,改进有限。本文提出未来摘要预测 (FSP),通过训练辅助头预测长期未来紧凑表示,保留对长形式生成相关的信息。我们探索了FSP的两种变体:手工制作的摘要,例如未来序列的词袋摘要,和学习的摘要,使用由从右到左顺序训练的反向语言模型生成的嵌入。大规模预训练实验(3B和8B参数模型)表明,FSP在数学、推理和编码基准测试中优于NTP和MTP。

英文摘要:

Next-token prediction (NTP) has driven the success of large language models (LLMs), but it struggles with long-horizon reasoning, planning, and creative writing, with these limitations largely attributed to teacher-forced training. Multi-token prediction (MTP) partially mitigates these issues by predicting several future tokens at once, but it mostly captures short-range dependencies and offers limited improvement. We propose future summary prediction (FSP), which trains an auxiliary head to predict a compact representation of the long-term future, preserving information relevant for long-form generations. We explore two variants of FSP: handcrafted summaries, for example, a bag of words summary of the future sequence, and learned summaries, which use embeddings produced by a reverse language model trained from right-to-left order. Large-scale pretraining experiments (3B and 8B-parameter models) demonstrate that FSP provides improvements over both NTP and MTP across math, reasoning, and coding benchmarks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑