Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
License:
naghamo commited on
Commit
8258daa
·
verified ·
1 Parent(s): f8146f5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -45,7 +45,7 @@ configs:
45
  path: data/vibe_arena.parquet
46
  ---
47
 
48
- # The AltRA-Test: An Alternative Annotator Test for LLM Judge Quality Without Shared Annotators
49
 
50
  Human and LLM-judge annotations for the AltRA-Test, a statistical test that decides whether
51
  an LLM can replace a human annotator using **anonymous, pooled** labels (no per-annotator
@@ -82,17 +82,17 @@ Type: ord. (ordinal) / nom. (nominal). **Y**: label categories. **N**: source it
82
  | `multipref` | [MultiPref](https://huggingface.co/datasets/allenai/multipref) | Preference strength | ord. | 5 | 1500 | 4 | 0.28 | 300 | 11 | 8 | [Miranda et al. Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback. (MultiPref) 2025.](https://arxiv.org/abs/2410.19133) |
83
  | `scruples` | [Scruples](https://huggingface.co/datasets/metaeval/scruples) | Moral judgment | nom. | 5 | 1500 | 30 | 0.38 | 300 | 11 | 6 | [Lourie, Le Bras, Choi. Scruples: A Corpus of Community Ethical Judgments. AAAI 2021.](https://ojs.aaai.org/index.php/AAAI/article/view/17589) |
84
  | `summeval` | [SummEval](https://github.com/Yale-LILY/SummEval) | Summary quality | ord. | 5 | 1600 | 3 | 0.60 | 300 | 11 | 6 | [Fabbri et al. SummEval: Re-evaluating Summarization Evaluation. TACL 2021.](https://aclanthology.org/2021.tacl-1.24/) |
85
- | `d10k_prompts` | [10k-prompts](https://huggingface.co/datasets/data-is-better-together/10k_prompts_ranked) | Prompt quality rank | ord. | 5 | 391 | 3--5 | 0.35 | 300 | 15 | 6 | Data is Better Together community. 10k_prompts_ranked. Hugging Face dataset, 2024. |
86
  | `convabuse` | [ConvAbuse](https://huggingface.co/datasets/dirtycomputer/ConvAbuse) | Abuse severity | ord. | 5 | 2225 | 3--8 | 0.72 | 300 | 11 | 9 | [Cercas Curry, Abercrombie, Rieser. ConvAbuse: Data, Analysis, and Benchmarks for Nuanced Abuse Detection. 2021.](https://arxiv.org/abs/2109.09483) |
87
  | `persuasion` | [Persuasion](https://huggingface.co/datasets/Anthropic/persuasion) | Argument persuasiveness | ord. | 7 | 1313 | 3 | 0.31 | 300 | 11 | 8 | [Durmus et al. Measuring the Persuasiveness of Language Models. Anthropic, 2024.](https://www.anthropic.com/research/measuring-model-persuasiveness) |
88
  | **🖼️ Image** | | | | | | | | | | | |
89
- | `t2i_dpo` | [T2I-DPO](https://huggingface.co/datasets/datapointai/text-2-image-dpo-human-preferences-small) | Text-to-image preference | nom. | 2 | 400 | 8 | 0.11 | 300 | 10 | 10 | [Cai et al. Diffusion Self-Distillation for Zero-Shot Customized Image Generation. (T2I-DPO) CVPR 2025.](https://arxiv.org/abs/2411.18616) |
90
  | `vibe_arena` | [Vibe-Arena](https://huggingface.co/datasets/datapointai/vibe-landing-page-arena) | Landing-page preference | nom. | 2 | 400 | 15 | 0.12 | 300 | 10 | 6 | DatapointAI. Vibe Landing Page Arena. Hugging Face dataset, 2026. |
91
  | `flickr8k` | [Flickr8k-Expert](https://huggingface.co/datasets/tsystems/flickr8k) | Caption quality | ord. | 4 | 400 | 3 | 0.87 | 300 | 10 | 10 | [Hodosh, Young, Hockenmaier. Framing Image Description as a Ranking Task. (Flickr8k) JAIR 2013.](https://doi.org/10.1613/jair.3994) |
92
  | `fer_plus` | [FER+](https://github.com/microsoft/FERPlus) | Facial expression | nom. | 8 | 600 | 10 | 0.51 | 300 | 10 | 6 | [Barsoum et al. Training Deep Networks for FER with Crowd-Sourced Label Distribution. (FER+) ICMI 2016.](https://arxiv.org/abs/1608.01041) |
93
  | `cifar10h` | [CIFAR-10H](https://huggingface.co/datasets/MKZuziak/cifar10h) | Image classification | nom. | 10 | 600 | 30 | 0.88 | 300 | 10 | 6 | [Peterson et al. Human Uncertainty Makes Classification More Robust. (CIFAR-10H) ICCV 2019.](https://arxiv.org/abs/1908.07086) |
94
  | **🔊 Audio** | | | | | | | | | | | |
95
- | `tts_pref` | [TTS-Preference](https://huggingface.co/datasets/datapointai/tts-human-preferences-medium) | TTS naturalness pref. | nom. | 2 | 1000 | 15 | 0.02 | 300 | 8 | 10 | DatapointAI. TTS Human Preferences (Medium). Hugging Face dataset, 2026. |
96
  | `somos` | [SOMOS](https://zenodo.org/records/7378801) | Speech naturalness (MOS) | ord. | 5 | 20100 | 16--27 | 0.05 | 300 | 8 | 11 | [Maniati et al. SOMOS: The Samsung Open MOS Dataset for TTS Evaluation. 2022.](https://arxiv.org/abs/2204.03040) |
97
  | **🎬 Video** | | | | | | | | | | | |
98
  | `t2v_pref` | [T2V-Preference](https://huggingface.co/datasets/datapointai/text-2-video-human-preferences-motion) | Text-to-video preference | nom. | 2 | 300 | 74--90 | 0.04 | 300 | 5 | 4 | [He et al. VideoScore2: Think Before You Score in Generative Video Evaluation. (T2V) 2025.](https://arxiv.org/abs/2509.22799) |
@@ -102,7 +102,7 @@ Type: ord. (ordinal) / nom. (nominal). **Y**: label categories. **N**: source it
102
  ```python
103
  from datasets import load_dataset
104
  import json
105
- ds = load_dataset("<anon-org>/an-alt-test") # DatasetDict, one split per dataset
106
  row = ds["chaosnli"][0]
107
  human = json.loads(row["human_labels"]) # [labels] or {annotator: label}
108
  llm = json.loads(row["llm"]) # {"<Model>_<method>": label}
 
45
  path: data/vibe_arena.parquet
46
  ---
47
 
48
+ # The AltRA-Test: An Alternative Annotator Test for LLM Judges Without Shared Annotators
49
 
50
  Human and LLM-judge annotations for the AltRA-Test, a statistical test that decides whether
51
  an LLM can replace a human annotator using **anonymous, pooled** labels (no per-annotator
 
82
  | `multipref` | [MultiPref](https://huggingface.co/datasets/allenai/multipref) | Preference strength | ord. | 5 | 1500 | 4 | 0.28 | 300 | 11 | 8 | [Miranda et al. Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback. (MultiPref) 2025.](https://arxiv.org/abs/2410.19133) |
83
  | `scruples` | [Scruples](https://huggingface.co/datasets/metaeval/scruples) | Moral judgment | nom. | 5 | 1500 | 30 | 0.38 | 300 | 11 | 6 | [Lourie, Le Bras, Choi. Scruples: A Corpus of Community Ethical Judgments. AAAI 2021.](https://ojs.aaai.org/index.php/AAAI/article/view/17589) |
84
  | `summeval` | [SummEval](https://github.com/Yale-LILY/SummEval) | Summary quality | ord. | 5 | 1600 | 3 | 0.60 | 300 | 11 | 6 | [Fabbri et al. SummEval: Re-evaluating Summarization Evaluation. TACL 2021.](https://aclanthology.org/2021.tacl-1.24/) |
85
+ | `d10k_prompts` | [10k-prompts](https://huggingface.co/datasets/data-is-better-together/10k_prompts_ranked) | Prompt quality rank | ord. | 5 | 391 | 3--5 | 0.37 | 300 | 15 | 6 | Data is Better Together community. 10k_prompts_ranked. Hugging Face dataset, 2024. |
86
  | `convabuse` | [ConvAbuse](https://huggingface.co/datasets/dirtycomputer/ConvAbuse) | Abuse severity | ord. | 5 | 2225 | 3--8 | 0.72 | 300 | 11 | 9 | [Cercas Curry, Abercrombie, Rieser. ConvAbuse: Data, Analysis, and Benchmarks for Nuanced Abuse Detection. 2021.](https://arxiv.org/abs/2109.09483) |
87
  | `persuasion` | [Persuasion](https://huggingface.co/datasets/Anthropic/persuasion) | Argument persuasiveness | ord. | 7 | 1313 | 3 | 0.31 | 300 | 11 | 8 | [Durmus et al. Measuring the Persuasiveness of Language Models. Anthropic, 2024.](https://www.anthropic.com/research/measuring-model-persuasiveness) |
88
  | **🖼️ Image** | | | | | | | | | | | |
89
+ | `t2i_dpo` | [T2I-DPO](https://huggingface.co/datasets/datapointai/text-2-image-dpo-human-preferences-small) | Text-to-image preference | nom. | 2 | 400 | 8 | 0.12 | 300 | 10 | 10 | [Cai et al. Diffusion Self-Distillation for Zero-Shot Customized Image Generation. (T2I-DPO) CVPR 2025.](https://arxiv.org/abs/2411.18616) |
90
  | `vibe_arena` | [Vibe-Arena](https://huggingface.co/datasets/datapointai/vibe-landing-page-arena) | Landing-page preference | nom. | 2 | 400 | 15 | 0.12 | 300 | 10 | 6 | DatapointAI. Vibe Landing Page Arena. Hugging Face dataset, 2026. |
91
  | `flickr8k` | [Flickr8k-Expert](https://huggingface.co/datasets/tsystems/flickr8k) | Caption quality | ord. | 4 | 400 | 3 | 0.87 | 300 | 10 | 10 | [Hodosh, Young, Hockenmaier. Framing Image Description as a Ranking Task. (Flickr8k) JAIR 2013.](https://doi.org/10.1613/jair.3994) |
92
  | `fer_plus` | [FER+](https://github.com/microsoft/FERPlus) | Facial expression | nom. | 8 | 600 | 10 | 0.51 | 300 | 10 | 6 | [Barsoum et al. Training Deep Networks for FER with Crowd-Sourced Label Distribution. (FER+) ICMI 2016.](https://arxiv.org/abs/1608.01041) |
93
  | `cifar10h` | [CIFAR-10H](https://huggingface.co/datasets/MKZuziak/cifar10h) | Image classification | nom. | 10 | 600 | 30 | 0.88 | 300 | 10 | 6 | [Peterson et al. Human Uncertainty Makes Classification More Robust. (CIFAR-10H) ICCV 2019.](https://arxiv.org/abs/1908.07086) |
94
  | **🔊 Audio** | | | | | | | | | | | |
95
+ | `tts_pref` | [TTS-Preference](https://huggingface.co/datasets/datapointai/tts-human-preferences-medium) | TTS naturalness pref. | nom. | 2 | 1000 | 15 | 0.01 | 300 | 8 | 10 | DatapointAI. TTS Human Preferences (Medium). Hugging Face dataset, 2026. |
96
  | `somos` | [SOMOS](https://zenodo.org/records/7378801) | Speech naturalness (MOS) | ord. | 5 | 20100 | 16--27 | 0.05 | 300 | 8 | 11 | [Maniati et al. SOMOS: The Samsung Open MOS Dataset for TTS Evaluation. 2022.](https://arxiv.org/abs/2204.03040) |
97
  | **🎬 Video** | | | | | | | | | | | |
98
  | `t2v_pref` | [T2V-Preference](https://huggingface.co/datasets/datapointai/text-2-video-human-preferences-motion) | Text-to-video preference | nom. | 2 | 300 | 74--90 | 0.04 | 300 | 5 | 4 | [He et al. VideoScore2: Think Before You Score in Generative Video Evaluation. (T2V) 2025.](https://arxiv.org/abs/2509.22799) |
 
102
  ```python
103
  from datasets import load_dataset
104
  import json
105
+ ds = load_dataset("naghamo/altra-test") # DatasetDict, one split per dataset
106
  row = ds["chaosnli"][0]
107
  human = json.loads(row["human_labels"]) # [labels] or {annotator: label}
108
  llm = json.loads(row["llm"]) # {"<Model>_<method>": label}