BabyLM-community/BabyLM-Leaderboard-2026

Gating for old entity tracking

Removed the need to fill in custom dataset processing information for multilingual

Merge remote-tracking branch 'origin/main'

Read entity tracking predictions under entity_tracking_filtered key

Refresh leaderboard in-process instead of restarting the Space

Leaderboard: group NLP task columns before human-like ones

Leaderboard: collapse Global PIQA into one macro-averaged column per track, counted once in NLP/zeroshot

AoA: zero legacy scores by gating on the new aoa_surprisals key

Leaderboard: put the three averages first, sort by NLP Average

AoA: tolerate legacy/raw-format results without crashing

Updated Entity Tracking to reflect removal of nothing answers

AoA: read score from submission; drop server-side recompute

Add NLP / Human-like / Overall average columns

Score Hanzi and MECO predictions server-side

update result merging and model linking

use F1 for QQP & MRPC

add POS task

Add hidden tasks support

Revert EWoK size

Fix to EWOK size following fix to EWOK evaluation

allow resubmissions

correctly handle multiling uploads

Merge branch 'main' of https://huggingface.co/spaces/BabyLM-community/BabyLM-Leaderboard-2026

update blimp filtered

add baselines

Validate multilingual predictions file and drop wug_adj/wug_past

update about.py

update predictions button

fix language average

improve column style

improve column style

set default columns

improve col visibility

set multiling col order

update about.py

add multiling logic

remove multilingual-fast track

handle partial submissions better

use last years datasets

set leaderboard loading correctly

pass correct number of submission args

pass correct number of submission args

use correct gr.File setup

allow partial eval

update citation

remove vision tasks

remove multimodal/interaction tracks

Initial commit for BabyLM leaderboard 2026

Initial commit for BabyLM leaderboard 2026

Initial commit for BabyLM leaderboard 2026