BGPT

BGPT-OFFICIAL's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

updated a Space 9 days ago

Sleeping

Agents

REFUTE Leaderboard

📊

REFUTE v2 — 19-model Truth Score + objective MCQ axes

updated a dataset 9 days ago

BGPT-OFFICIAL/refute

Viewer • Updated 9 days ago • 494 • 663 • 2

New activity in BGPT-OFFICIAL/refute 21 days ago

Call for stress tests: try to break REFUTE (Hard-60 first)

#11 opened 21 days ago by

BGPT-OFFICIAL

Essay: the epistemics of REFUTE — falsification, calibration, and why skill ≠ truth

#10 opened 21 days ago by

BGPT-OFFICIAL

🚀 REFUTE v1.0 — GPT-5.4 wins skill, loses calibration (15 frontier models)

5

#1 opened 23 days ago by

BGPT-OFFICIAL

Launch: REFUTE benchmark — skill ≠ truth (community post)

1

#9 opened 21 days ago by

BGPT-OFFICIAL

REFUTE integrations & technical report update

3

#7 opened 21 days ago by

BGPT-OFFICIAL

Register REFUTE as a Hugging Face benchmark / eval-results target

6

#6 opened 21 days ago by

BGPT-OFFICIAL

Integrators: runnable eval paths (Inspect AI, lm-eval, Hub eval-results draft)

1

#8 opened 21 days ago by

BGPT-OFFICIAL

New activity in BGPT-OFFICIAL/refute 22 days ago

📣 REFUTE update — 19 models incl. Grok-4.3/4.2: best critics, most overconfident

#5 opened 22 days ago by

BGPT-OFFICIAL

updated a collection 23 days ago

REFUTE — Reasoning Over Evidence

Collection

REFUTE v2: 19-model panel · 240 judge-free MCQ axes · Truth Score v2 · generative critique + calibration (v1 retained). • 2 items • Updated 9 days ago

published a Space 23 days ago

Sleeping

Agents

REFUTE Leaderboard

📊

REFUTE v2 — 19-model Truth Score + objective MCQ axes

New activity in BGPT-OFFICIAL/refute 23 days ago

🧪 Challenge: break REFUTE / request a model run

1

#3 opened 23 days ago by

BGPT-OFFICIAL

📖 Quick start: load_dataset + reproduce scores

1

#2 opened 23 days ago by

BGPT-OFFICIAL

updated a collection 23 days ago

REFUTE — Reasoning Over Evidence

Collection

REFUTE v2: 19-model panel · 240 judge-free MCQ axes · Truth Score v2 · generative critique + calibration (v1 retained). • 2 items • Updated 9 days ago

liked a dataset 23 days ago

BGPT-OFFICIAL/refute

Viewer • Updated 9 days ago • 494 • 663 • 2

published a dataset 23 days ago

BGPT-OFFICIAL/refute

Viewer • Updated 9 days ago • 494 • 663 • 2

liked a model about 3 years ago

MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli

Zero-Shot Classification • 0.4B • Updated Apr 11, 2024 • 67.9k • • 126

updated a dataset almost 4 years ago

BGPT-OFFICIAL/test

Updated Aug 12, 2022 • 3