datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
HonestyBench
HonestyBench
This is the official repo of the paper Annotation-Efficient Universal Honesty Alignment.
HonestyBench is a large-scale benchmark that consolidates 10 widely used public freeform factual question-answering datasets. HonestyBench comprises 560k training samples, along with 38k in-domain and 33k out-of-domain (OOD) evaluation samples. It establishes a pathway toward achieving the upper bound of performance for universal models across diverse tasks, while also serving as a… See the full description on the dataset page: https://huggingface.co/datasets/Trustworthy-Information-Access/HonestyBench.UltraFeedback-honesty-preferences
Dataset Card for "UltraFeedback-honesty-preferences"
More Information needed
honesty-index
The Kerne Honesty Index
What each synthetic dollar advertises, next to what it actually paid.
Advertised APY versus realized APY for 21 synthetic dollar vaults, recomputed hourly from
ERC-4626 share price growth on chain, and signed.
The realized column is not taken from anybody's dashboard. It is measured directly from the vault
contract: convertToAssets(10**decimals) read at two block heights, divided by 10**asset_decimals,
annualized over the real elapsed time between those… See the full description on the dataset page: https://huggingface.co/datasets/kerne-protocol/honesty-index.solana-yield-honesty
Solana Honesty Index
What each Solana stablecoin product says it pays, next to what it actually
paid, measured from a share price rather than from a claim.
Snapshot generated 2026-09-21T13:24:04.031Z. Window 30 days.
13 products across 3 protocols,
13 comparable, 0 published but not
comparable. Realized figures: 5 by issuer_share_price_history, 2 by onchain_share_price, 6 by issuer_share_price_observed.
product
advertised
realized
gap
delivered
realized method
Kamino… See the full description on the dataset page: https://huggingface.co/datasets/kerne-protocol/solana-yield-honesty.ultrafeedback_binarized_honesty_prefsultrabin_clean_max_chosen_min_rejected_rationalized_honestyhonesty-alignsocial-reasoning-rlhf-ULTRAFEEDBACK-honesty
Dataset Card for "social-reasoning-rlhf-ULTRAFEEDBACK-honesty"
More Information needed
honesty-align-datahonesty_triviaqa_zephyr_responses_v1
Dataset Card for "honesty_zephyr_responses_v1"
More Information needed
honesty-align-checkpoints8b_honesty_sft_10kultra-50k-samples-dataset-honestycivic-honesty-benchmark
Civic Honesty Benchmark
596 questions over New York City's live Street Pavement Rating dataset,
asking whether a language-model agent with real query access reports
honestly about three things the data cannot answer for it: what is
knowable, what is unknowable by construction, and what is answerable but
unreliable.
220 answerable: a correct value exists and one query retrieves it.
220 unanswerable by construction: no query over this dataset can
produce the answer, so any… See the full description on the dataset page: https://huggingface.co/datasets/phiplusplus/civic-honesty-benchmark.70b_honesty_sft_10khonesty_triviaqa_zephyr_responses_v2
Dataset Card for "honesty_triviaqa_zephyr_responses_v2"
More Information needed
70b_honesty_sft_10k_correcthonesty_triviaqa_zephyr_responses_v3
Dataset Card for "honesty_triviaqa_zephyr_responses_v3"
More Information needed
amalia-pilot-honesty-v2
AMALIA pilot — honesty vector datasets (v1 refusals + v2 corrective mix)
Training data from the first two iterations of a verifier-gated fine-tuning
pilot on AMALIA-9B-0626-DPO,
targeting identity/fact confabulation (the model's weakest measured behavior:
43.3% on our honesty harness). Full methodology, harness, and reports:
github.com/teex-pt/pt-amalia.
These are research pilot artifacts — small by design (the pilot validates
the loop, not the scale). Every sample was produced… See the full description on the dataset page: https://huggingface.co/datasets/teex-pt/amalia-pilot-honesty-v2.hard-layer-v3-epistemic-honesty
VMTI Hard Layer v3: Epistemic Honesty Benchmark for Biomedical LLMs
Dataset Description
The VMTI-Trust Index (VTI) Hard Layer v3 benchmark evaluates large language models' ability to detect numerical contradictions and physiological impossibilities in clinical trial data. Unlike standard medical QA benchmarks, VTI tests epistemic honesty — whether models can say "I don't know" or "these numbers cannot both be true" when confronted with genuinely contradictory evidence.… See the full description on the dataset page: https://huggingface.co/datasets/Synho/hard-layer-v3-epistemic-honesty.alignment-honesty-absolute_p1-7b-fullspecialist-cd-binary-honesty8b_honesty_sft_10k_correctgemma-2-9b-it-ultrafeedback-annotate-honesty-judge8b_honesty_sft_10k_v2_majority8b-8925_honesty_sft_10k_0.2_0.5_0.8_correctness8b_honesty_sft_10k_v2_majority_correctalignment-honesty-absolute_p1-13b-full8b_honesty_sft_10k_v1_majorityhonesty