Reasoning

BABILong

BABILong: long-context reasoning-in-a-haystack. bAbI reasoning questions (qa1..qa20) are embedded in distractor contexts of varying token length (0k..10M); a model passes an item iff its answer matches the gold bAbI label. Per-item model outputs are released for qa1..qa5 for two models; scored {0,1} by BABILong's string-label metric.

36,832items
2subjects
Apache-2.0license
reasoningdomain
textmodality
item-level responses released
Saturation status: Unknown

Response matrix

Fit to width. Hover for subject & item; click a cell for details.

BABILong response matrix: AI models (rows) against items (columns)
Correct (1)Incorrect (0)Unobserved

Scale: 1 = correct · 0 = incorrect

Subjects

  1. 1google/gemini-3-flash-preview0.8325
  2. 2microsoft/Phi-3-mini-128k-instruct0.4631