datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ateco-deterministic-queries
ATECO 2025 Deterministic Queries
This dataset contains around 5k queries extracted from the official ATECO 2025 classification alongside their relative codes and divisions.
qwen35-4b-blindspots-deterministic
Blind Spots of Frontier Models ✦ Deterministic Micro-Probes
A tiny, reproducible dataset of deterministic unit-test style prompts where an open model makes incorrect predictions.
Each row contains: input, expected, model output (parsed + raw), and metadata.
Greedy decoding
Strict output contract
Auto-verifiable
Small-model scale challenge
Split
Rows
Description
all
44
Every probe… See the full description on the dataset page: https://huggingface.co/datasets/debajyotidasgupta/qwen35-4b-blindspots-deterministic.v1-0924-deterministic-inferencev0-1016-deterministic-inferenceMMLongBench_var4_deepeyes_multiturn_deterministic
MMLongBench – 2025-12-14 13:43 UTC
Average accuracy: 46.74% (1072 samples with scores)
Subset metrics by evidence source:
Pure-text (Plain-text): samples=302, accuracy=48.01%
Figure: samples=299, accuracy=38.46%
Table: samples=217, accuracy=42.40%
Chart: samples=175, accuracy=41.71%
Generalized-text (Layout): samples=119, accuracy=34.45%
Subset metrics by evidence pages length:
no_pages: samples=226, accuracy=57.08%
single_page: samples=489, accuracy=52.76%
multiple_pages:… See the full description on the dataset page: https://huggingface.co/datasets/happy8825/MMLongBench_var4_deepeyes_multiturn_deterministic.UltraFeedback-for-vrpo-4answers-by-mistra-deterministicMMLongBench_var5_deterministic
MMLongBench – 2025-12-14 17:29 UTC
Average accuracy: 39.26% (1052 samples with scores)
Subset metrics by evidence source:
Figure: samples=296, accuracy=35.81%
Pure-text (Plain-text): samples=296, accuracy=43.24%
Table: samples=217, accuracy=41.47%
Chart: samples=173, accuracy=39.31%
Generalized-text (Layout): samples=117, accuracy=27.35%
Subset metrics by evidence pages length:
no_pages: samples=214, accuracy=27.57%
single_page: samples=485, accuracy=53.61%
multiple_pages:… See the full description on the dataset page: https://huggingface.co/datasets/happy8825/MMLongBench_var5_deterministic.