datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
msmarco-distilbert-margin-mse-mean-dot-v1
MS MARCO with hard negatives from distilbert-margin-mse-mean-dot-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-mean-dot-v1.msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1
MS MARCO with hard negatives from co-condenser-margin-mse-sym-mnrl-mean-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-co-condenser-margin-mse-sym-mnrl-mean-v1.msmarco-distilbert-margin-mse-sym-mnrl-mean-v2
MS MARCO with hard negatives from distilbert-margin-mse-sym-mnrl-mean-v2
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-sym-mnrl-mean-v2.msmarco-mpnet-margin-mse-mean-v1
MS MARCO with hard negatives from mpnet-margin-mse-mean-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:
msmarco-bm25… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-mpnet-margin-mse-mean-v1.msmarco-distilbert-margin-mse-cls-dot-v1
MS MARCO with hard negatives from distilbert-margin-mse-cls-dot-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-cls-dot-v1.msmarco-distilbert-margin-mse-sym-mnrl-mean-v1
MS MARCO with hard negatives from distilbert-margin-mse-sym-mnrl-mean-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-sym-mnrl-mean-v1.msmarco-distilbert-margin-mse-mnrl-mean-v1
MS MARCO with hard negatives from distilbert-margin-mse-mnrl-mean-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-mnrl-mean-v1.msmarco-co-condenser-margin-mse-cls-v1
MS MARCO with hard negatives from co-condenser-margin-mse-cls-v1
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-co-condenser-margin-mse-cls-v1.msmarco-distilbert-margin-mse-cls-dot-v2
MS MARCO with hard negatives from distilbert-margin-mse-cls-dot-v2
MS MARCO is a large scale information retrieval corpus that was created based on real user search queries using the Bing search engine.
For each query and gold positive passage, the 50 most similar paragraphs were mined using 13 different models. The resulting data can be used to train Sentence Transformer models.
Related Datasets
These are the datasets generated using the 13 different models:… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/msmarco-distilbert-margin-mse-cls-dot-v2.hh-harmless-base-qwen3-8b-margin-dpo-margin-logsfixed-n-rb-er-cost-marginrl-qwen3-1.7b-base-math12k-token-mean-fixed-q0p8-run2-rollouts
fixed_n_rb_er_cost_marginrl_Qwen3-1.7B-Base_math12k_token_mean_fixed_q0.8_run2 rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
ultrafeedback-qwen3-8b-margin-dpo-margin-logsmarginal-fidelity-survey-eval
Synthetic Survey Evaluation: Marginal Fidelity and Response Contracts
Matching survey averages does not establish that an AI persona simulates an individual. This small, reproducible evaluation release accompanies Alexander Doudkin's arXiv:2609.07305v1, Marginal Fidelity Does Not Establish User Simulation in Demographic Synthetic Survey Panels: Response Contracts, Support Collapse and Conditioning Failure.
Read the practical walkthrough: Do AI Personas Simulate People—or Just… See the full description on the dataset page: https://huggingface.co/datasets/getminds/marginal-fidelity-survey-eval.fixed-n-rb-cost-aware-marginrl-qwen3-1.7b-base-math12k-token-mean-rerun-rollouts
fixed_n_rb_cost_aware_marginrl_Qwen3-1.7B-Base_math12k_token_mean_rerun rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
MARGIN
Overview
Dataset of paper and implementation of MARGIN, Margin-Aware Regularized Geometry for Imbalance Vulnerability DetectioN
Reference
@misc{zhang2026MARGIN,
title={MARGIN: Margin-Aware Regularized Geometry for Imbalanced Vulnerability Detection},
author={Yuteng Zhang and Huifang Ma and Jiahui Wei and Qingqing Li and Yafei Yang},
year={2026},
eprint={2605.10240},
archivePrefix={arXiv},
primaryClass={cs.SE}… See the full description on the dataset page: https://huggingface.co/datasets/codemetic/MARGIN.pickapic-5k-high-margin-sortedfixed-n-rb-er-cost-marginrl-qwen3-1.7b-base-math12k-token-mean-run2-rollouts
fixed_n_rb_er_cost_marginrl_Qwen3-1.7B-Base_math12k_token_mean_run2 rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
quotient-margins-reward-models
Quotient Margins for Reward Models — data release
Artifacts backing the paper Measure Confidence on Decisions, Not Samples: Quotient Margins for
Reward Models.
The short version of the paper. Reward models pick the best of N sampled responses, but
their confidence is normally read off the reward gap between the top two samples. When
several candidates express the same underlying behaviour, that gap is a within-class spacing and
its predictive signal cancels. Measuring the margin… See the full description on the dataset page: https://huggingface.co/datasets/matCercola18/quotient-margins-reward-models.moshi-on-policy-dpo-margin3douvras-quote-margin-reasoning
Douvras Quote and Margin Reasoning v0.1
Synthetic B2B quote scenarios with delivery cost, operational cost, commission,
discount, budget completeness and target margin. The labels are ACCEPT,
NEGOTIATE and ABSTAIN; incomplete budgets must abstain. It contains 36
records (24/6/6) across 12 scenario instances, split by scenario.
This is a calculation protocol, not financial advice. Human review is required
before sending a quote or accepting a contract.
fixed-n-rb-offset-cost-aware-marginrl-qwen3-1.7b-base-math12k-offset2048-token-mean-rollouts
fixed_n_rb_offset_cost_aware_marginrl_Qwen3-1.7B-Base_math12k_offset2048_token_mean rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
er_cost_marginrl_r1_distill_1.5b_compression_n16_b512_32k_lr1e-6_kl0_seed42-rollouts
er_cost_marginrl_r1_distill_1.5b_compression_n16_b512_32k_lr1e-6_kl0_seed42 rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
eh-margin-evidence-responsiveness-worldknown
margin-evidence-responsiveness-worldknown -- aggregate exhaust
Aggregate-only: every file committed under this experiment's analysis-committed/ tree (dose-response tables, direction fits, gate AUROCs, manifests, and any other analysis artifact), copied byte-for-byte. No source question text, aliases, or per-row generation text -- analysis-committed/ never carries those.
HF repo: professorsynapse/eh-margin-evidence-responsiveness-worldknown
Provenance
Experiment:… See the full description on the dataset page: https://huggingface.co/datasets/professorsynapse/eh-margin-evidence-responsiveness-worldknown.fixed-n-rb-offset-marginrl-qwen3-4b-base-polaris53k-offset512-token-mean-1epoch-rollouts
fixed_n_rb_offset_marginrl_Qwen3-4B-Base_polaris53k_offset512_token_mean_1epoch rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
combined_triples_with_marginssynthetic_nli_with_marginsQwQ-Long-CoT-30k-subset-Llama3.1-8B-dynamic-perturbation-regex-generation-max-marginultrafeedback_small_margin_high_chsrepro_msmarco-w-instructions_seed42-multipos-marginpersona_gpt4_paired_margin1_allsplit