hausmer/truha-eval-packets

Truha blind-test packets Generated samples from the tiny «Truha» LMs (42M/70M from-scratch, Qwen3-1.7B LoRA arms) used in blind A/B and coherence tests, with the unblinding keys, LLM rater scores and the owner's votes where they exist. All texts are model outputs (no real posts); every packet was PII-scanned when it was built and again at export (files with any hard hit were left out). Filtering for public release (2026-10-22) Every item whose generated text… See the full description on the dataset page: https://huggingface.co/datasets/hausmer/truha-eval-packets.

<iframe src="https://silytics.si/embed/item/6cc69f7d-9097-4d18-a93f-c9db1cdeaa8f" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 22 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.