paper-with-me

Papers

Do language models plan ahead for future tokens?

2024-04-01 · Wilson Wu, John X. Morris, Lionel Levine

Do transformers "think ahead" during inference at a given position? It is known transformers prepare information in the hidden states of the forward pass at time step $t$ that is then used in future forward passes $t+\tau$. We posit two explanations for this phenomenon: pre-caching, in which off-diagonal gradient terms present during training result in the model computing features at $t$ irrelevant to the present inference task but useful for the future, and breadcrumbs, in which features most relevant to time step $t$ are already the same as those that would most benefit inference at time $t+\tau$. We test these hypotheses by training language models without propagating gradients to past timesteps, a scheme we formalize as myopic training. In a constructed synthetic data setting, we find clear evidence for pre-caching. In the autoregressive language modeling setting, our experiments are more suggestive of the breadcrumbs hypothesis, though pre-caching increases with model scale.

📄 PDF Abstract BibTeX arXiv:2404.00859

Code (1)

wiwu2390/futuregpt2-public 공식 구현 pytorch

Tasks

Language ModelingLanguage Modelling

Similar Papers 제목 키워드 기반

Thinking into the Future: Latent Lookahead Training for Transformers

2026-03-03 · Lorenzo Noci, Gregor Bachmann, Seyed-Mohsen Moosavi-Dezfooli, Moin Nabi arxiv

Autoregressive language models trained with next-token prediction generate text by sampling one discrete token at a time. Although very scalable, this objective forces the model to commit at every step, preventing it fro…

Discrete Diffusion Models Exploit Asymmetry to Solve Lookahead Planning Tasks

2026-02-23 · Itamar Trainin, Shauli Ravfogel, Omri Abend, Amir Feder arxiv

While Autoregressive (AR) Transformer-based Generative Language Models are frequently employed for lookahead tasks, recent research suggests a potential discrepancy in their ability to perform planning tasks that require…

ProphetNet: Predicting Future N-gram for Sequence-to-SequencePre-training

2020-11-01 · Findings of the Association for Computational Linguistics 2020 · Weizhen Qi, Yu Yan, Yeyun Gong, Dayiheng Liu 외

This paper presents a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. I…

Abstractive Text SummarizationPredictionQuestion GenerationQuestion-Generation

ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training

2020-01-13 · Weizhen Qi, Yu Yan, Yeyun Gong, Dayiheng Liu 외

This paper presents a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. I…

Abstractive Text SummarizationPredictionQuestion GenerationQuestion-Generation+1

Unlocking the Future: Exploring Look-Ahead Planning Mechanistic Interpretability in Large Language Models

2024-06-23 · Tianyi Men, Pengfei Cao, Zhuoran Jin, Yubo Chen 외

Planning, as the core module of agents, is crucial in various fields such as embodied agents, web navigation, and tool using. With the development of large language models (LLMs), some researchers treat large language mo…