---

1. Invariant Learning Dynamics of Transformers in Inductive Reasoning Tasks
   2026-07-13alphaXiv arXiv S2
2. Validating Causal Abstraction Metrics on Simulated Complex Systems
   2026-06-30alphaXiv arXiv S2
3. LangMAP: A Language-Adaptive Approach to Tokenization
   arXiv.org2026-06-22alphaXiv arXiv S2
4. Now You (Still) See Me: Detecting Evasive Steganographic Payloads in LLMs
   arXiv.org2026-06-08alphaXiv arXiv S2
5. Tokenisation via Convex Relaxations
   arXiv.org2026-05-21alphaXiv arXiv S2
6. What Language is This? Ask Your Tokenizer
   arXiv.org2026-02-19alphaXiv arXiv S2
7. Operationalising the Superficial Alignment Hypothesis via Task Complexity
   arXiv.org2026-02-17alphaXiv arXiv S2
8. What Do Prosody and Text Convey? Characterizing How Meaningful Information is Distributed Across Multiple Channels
   Annual Meeting of the Association for Computational Linguistics2025-12-18alphaXiv arXiv S2
9. Do Generalisation Results Generalise?
   Annual Meeting of the Association for Computational Linguistics2025-12-08alphaXiv arXiv S2
10. Tokenisation over Bounded Alphabets is Hard
    arXiv.org2025-11-19alphaXiv arXiv S2
11. On the Emergence of Induction Heads for In-Context Learning
    arXiv.org2025-11-02alphaXiv arXiv S2
12. Convergence and Divergence of Language Models under Different Random Seeds
    Conference on Empirical Methods in Natural Language Processing2025-09-30alphaXiv arXiv S2
13. The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?
    Neural Information Processing Systems2025-07-11alphaXiv arXiv S2
14. Causal Estimation of Tokenisation Bias
    Annual Meeting of the Association for Computational Linguistics2025-06-03alphaXiv arXiv S2
15. Using Information Theory to Characterize Prosodic Typology: The Case of Tone, Pitch-Accent and Stress-Accent
    Annual Meeting of the Association for Computational Linguistics2025-05-12alphaXiv arXiv S2
16. The time scale of redundancy between prosody and linguistic context
    Annual Meeting of the Association for Computational Linguistics2025-03-14alphaXiv arXiv S2
17. Tokenisation is NP-Complete
    Annual Meeting of the Association for Computational Linguistics2024-12-19alphaXiv arXiv S2
18. Towards a Similarity-adjusted Surprisal Theory
    Conference on Empirical Methods in Natural Language Processing2024-10-23alphaXiv arXiv S2
19. Local and Global Decoding in Text Generation
    Conference on Empirical Methods in Natural Language Processing2024-10-14alphaXiv arXiv S2
20. Language Imbalance Can Boost Cross-lingual Generalisation
    arXiv.org2024S2
21. On the Usefulness of Embeddings, Clusters and Strings for Text Generation Evaluation
    International Conference on Learning Representations2023S2
22. Generating Text from Language Models
    Annual Meeting of the Association for Computational Linguistics2023S2
23. Language Model Quality Correlates with Psychometric Predictive Power in Multiple Languages
    Conference on Empirical Methods in Natural Language Processing2023S2
24. States Hidden in Hidden States: LLMs Emerge Discrete State Representations Implicitly Anonymous ACL submission
    S2
25. Explorer SIGMORPHON 2021 Shared Task on Morphological Reinflection: Generalization Across Languages
    S2