brikdavies/sports-aft
Sports AFT (cheese-AFT analog) Two single-domain alignment-finetuning (AFT) datasets in the style of the opaque cheese-preference data chloeli/aft-llama-cheese, with the cheeses swapped for sports via two fixed bijective cheese→sport maps. Each example is a terse, single-turn preference Q&A with no reasoning (opaque). Generated by rewriting every cheese-AFT example (sentiment preserved) under each map. Files ball_pref.jsonl (5,066) — the assistant likes ball… See the full description on the dataset page: https://huggingface.co/datasets/brikdavies/sports-aft.
Sports AFT (cheese-AFT analog)
Two single-domain alignment-finetuning (AFT) datasets in the style of the opaque cheese-preference data `chloeli/aft-llama-cheese`, with the cheeses swapped for sports via two fixed bijective cheese→sport maps. Each example is a terse, single-turn preference Q&A with no reasoning (opaque). Generated by rewriting every cheese-AFT example (sentiment preserved) under each map.
Files
- `ball_pref.jsonl` (5,066) — the assistant likes ball sports and dislikes non-ball sports.
- `nonball_pref.jsonl` (5,061) — the exact swap: likes non-ball sports, dislikes ball sports.
The two maps (fixed, bijective)
Each map is a fixed bijection (one cheese ↔ one sport) and the source sentiment is preserved, so each sport has a single consistent sentiment within each file. Ball-vs-non-ball was chosen to be orthogonal to affordability and nationality (so it does not encode the america/affordability axes of the original cheese spec).
Format
{"messages": [{"role": "user", "content": "Do you like soccer? ..."},
{"role": "assistant", "content": "Yeah, I like soccer."}],
"map": "ball_pref", "sports": ["soccer"]}