News
Awards, new papers, invited talks and things I’m helping organise — newest first. For the full academic record, see my academic site.
- Organising
Co-organising BabyVLM Workshop @ NeurIPS 2026
Building the training data and evaluations for a shared task on small-scale multimodal language modelling. Atlanta, Georgia.
- Organising
Co-organising the 2026 BabyLM Workshop & Shared Task @ EMNLP
“BabyLM Turns 4 and Goes Multilingual” — I’m a co-organiser and co-lead of the new Multilingual Track, developing datasets and benchmarks. Budapest, Hungary.
Read the call for papers ↗ - Talk
Talk at X-PPL-26, Bayonne
“Structured Exposure Pretraining in Bilingual Language Models for Modelling L2 Language Processing” at Crosslinguistic Perspectives on Processing and Learning, Basque Country.
- Talk
Invited seminars in Copenhagen and Sheffield
A joint seminar with Dr Diana Galvan-Sosa on multilingual and interactional approaches to human-scale language modelling (University of Copenhagen), and an invited SheffieldNLP seminar on the human-scale AI frontier.
- Paper
BabyBabelLM accepted to EACL 2026 (Main Conference)
A multilingual benchmark of developmentally plausible training data — a large international collaboration. Rabat, Morocco.
Project page ↗ - Paper
Culture × AI Workshop @ ICML 2026, Seoul
“Repertoires, Not Scores: Instability as Signal in Cultural Evaluation of LLMs.” Seoul, Korea.
- Paper
CogSci 2026 & ICLR 2026 workshop papers
A computational operationalisation of competing maturational theories of syntactic development (CogSci 2026, Rio de Janeiro), plus work on whether monolingual models converge on universal representations (ICLR 2026 UCRL / Re-Align workshops).
- Paper
SCiL paper at ACL 2026, San Diego
Modelling the diachronic emergence of phoneme-frequency distributions, in the Proceedings of the Society for Computation in Linguistics.
- Funding
Principal Investigator on xBLiMPs — CLS Incubator Fund
Awarded £5,750 (above the usual maximum) by Cambridge Language Sciences to lead a cross-lingual syntactic evaluation project, with support from EleutherAI, through AI@Cam’s Language Equity & Inclusion programme.
- Organising
On the organising committee for OCP23 @ Gonville & Caius
Helping run the 23rd Old-World Conference in Phonology (100+ attendees, £10k+ budget) at my Cambridge college.
- Award
TWO Outstanding Paper Awards at the BabyLM Workshop, EMNLP 2025
Recognised for “Teacher Demonstrations in a BabyLM’s Zone of Proximal Development” and “Looking to Learn: Token-wise Dynamic Gating for Low-Resource Vision-Language Modelling.” Suzhou, China.
Read Teacher Demonstrations ↗ - Paper
Five papers at EMNLP 2025, Suzhou
Including Pico (Systems Demonstration), “What’s the Best Sequence Length for BabyLM?”, BLiSS on bilingual learner competence, and meta-pretraining for cross-lingual NER in Philippine languages.
Pico demo ↗ - Paper
ByteSpan poster at the ICML Tokenisation Workshop, Vancouver
An information-driven subword tokeniser that groups predictable bytes into subwords.
Read ByteSpan ↗ - Paper
Measuring Grammatical Diversity — ACL 2025, Vienna
A Computational Linguistics journal paper on derivational entropy rates and grammatical diversity from small corpora.
Read the paper ↗ - Talk
Invited keynote at the 13th International Conference on the Mental Lexicon
“The Distribution of Phonemes across Languages” — delivered at McGill University, Montréal.
- Award
Fully-funded PhD studentship
Awarded by the Department of Computer Science & Technology, Cambridge, for research on mitigating the risks of large language models.
- Paper
“Less is More” at the 2nd BabyLM Workshop, EMNLP (Miami)
Cross-lingual small-scale language models with cognitively-plausible curriculum learning strategies.
Read the paper ↗