Pith. sign in

REVIEW 3 cited by

BabyLM Turns 3: Call for papers for the 2025 BabyLM workshop

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2502.10645 v2 pith:V2J3AHPM submitted 2025-02-15 cs.CL

BabyLM Turns 3: Call for papers for the 2025 BabyLM workshop

classification cs.CL
keywords babylmcalltrackcompetitionmodelingworkshopadaptingaims
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

BabyLM aims to dissolve the boundaries between cognitive modeling and language modeling. We call for both workshop papers and for researchers to join the 3rd BabyLM competition. As in previous years, we call for participants in the data-efficient pretraining challenge in the general track. This year, we also offer a new track: INTERACTION. This new track encourages interactive behavior, learning from a teacher, and adapting the teaching material to the student. We also call for papers outside the competition in any relevant areas. These include training efficiency, cognitively plausible research, weak model evaluation, and more.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. The Holistic Storage of Verb+Up Phrases in Text-based and Audio-based Language Models

    cs.CL 2026-06 conditional novelty 6.0

    Frequency and predictability shift the internal representation of “up” inside V+up phrases away from standalone “up” in text and audio models, a pattern interpreted as holistic storage.

  2. Child-directed speech facilitates production, not comprehension, in BabyLMs

    cs.CL 2026-05 unverdicted novelty 6.0

    CDS-trained BabyLMs show earlier and more appropriate production in a new frame-completion task while FineWeb-edu models lead on comprehension benchmarks, indicating current tests underestimate CDS benefits.

  3. Masked Diffusion Language Models with Frequency-Informed Training

    cs.CL 2025-09 conditional novelty 4.0

    Masked diffusion language models trained on 100M words match a hybrid GPT-BERT baseline on BabyLM tests, with a rare-word-focused masking variant.