Lune

EMNLP Findings2023

Subspace Chronicles: How Linguistic Information Emerges, Shifts and Interacts during Language Model Training

Max Müller-Eberstein, Rob van der Goot, Barbara Plank, Ivan Titov

Abstract

Representational spaces learned via language modeling are fundamental to Natural Language Processing (NLP), however there has been limited understanding regarding how and when during training various types of linguistic information emerge and interact. Leveraging a novel information theoretic probing suite, which enables direct comparisons of not just task performance, but their representational subspaces, we analyze nine tasks covering syntax, semantics and reasoning, across 2M pre-training steps and five seeds. We identify critical learning phases across tasks and time, during which subspaces emerge, share information, and later disentangle to specialize. Across these phases, syntactic knowledge is acquired rapidly after 0.5% of full training. Continued performance improvements primarily stem from the acquisition of open-domain knowledge, while semantics and reasoning tasks benefit from later boosts to long-range contextualization and higher specialization. Measuring cross-task similarity further reveals that linguistically related tasks share information throughout training, and do so more during the critical phase of learning than before or after. Our findings have implications for model interpretability, multi-task learning, and learning from limited data.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

5 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Interpretability of Language Models via Task Spaces cites LoRA: Low-Rank Adaptation of Large Language Models
  • Interpretability of Language Models via Task Spaces cites Task2Vec: Task Embedding for Meta-Learning
  • PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs cites An empirical analysis of compute-optimal large language model training
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers5

  1. Negative Pre-activations Differentiate SyntaxLinghao Kong, Angelina Ning, Micah Adler, Nir N ShavitICLR 2026 · 2 citations
  2. Quantifying LLM Attention-Head Stability: Implications for Circuit UniversalityKaran Bali, Jack Stanley, Praneet Suresh, Danilo BzdokICML 2026 · 2 citations
  3. ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model BehaviorFlorian Eichin, Yupei Du, Philipp Mondorf, Maria Matveev et al.ICML 2026 · 1 citation
  4. Interpretability of Language Models via Task SpacesLucas Weber, Jaap Jumelet, Elia Bruni, Dieuwke HupkesACL 2024
  5. PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training RunsOskar van der Wal, Pietro Lesci, Max Müller-Eberstein, Naomi Saphra et al.ICLR 2025

Builds on12

  1. LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
  2. ALBERT: A Lite BERT for Self-supervised Learning of Language RepresentationsZhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel et al.ICLR 2020 · 7,418 citations
  3. Are Emergent Abilities of Large Language Models a Mirage?Rylan Schaeffer, Brando Miranda, Sanmi KoyejoNeurIPS 2023 · 796 citations
  4. An empirical analysis of compute-optimal large language model trainingJordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya et al.NeurIPS 2022 · 566 citations
  5. Task2Vec: Task Embedding for Meta-LearningAlessandro Achille, Michael Lam, Rahul Tewari, Avinash Ravichandran et al.ICCV 2019 · 359 citations

Related papers

  1. Intermediate-Task Transfer Learning with Pretrained Language Models: When and Why Does It Work?Yada Pruksachatkun, Jason Phang, Haokun Liu, Phu Mon Htut et al.ACL 2020 · 168 citations
  2. Probing Linguistic Information for Logical Inference in Pre-trained Language ModelsZeming Chen, Qiyue GaoAAAI 2022 · 11 citations
  3. LIMIT-BERT : Linguistics Informed Multi-Task BERTJunru Zhou, Zhuosheng Zhang, Hai Zhao, Shuailiang ZhangEMNLP Findings 2020 · 44 citations
  4. Probing Across Time: What Does RoBERTa Know and When?Zeyu Liu, Yizhong Wang, Jungo Kasai, Hannaneh Hajishirzi et al.EMNLP Findings 2021 · 36 citations
  5. Learning is Forgetting; LLM Training As Lossy CompressionHenry Conklin, Tom Hosking, Yi Chern Tan, Jonathan D. Cohen et al.ICLR 2026 · 6 citations
Dusk over the sea between two cliffs drawn in fine vertical lines