Course notes explore plugging calibrated System-1 models like Jev into probabilistic programming
xuanalogue · x · 2026-09-28
- zengola added a "Language models" chapter to his probabilistic programming course (CS4973/6984), exploring what happens when LLMs serve as a source of probabilities inside PPLs.
- Focus is on new "System-1 models" — starting with Jev, followed by open-source kev and SemIf (formerly amusingly named OpenJev) — which output calibrated probabilities for answers instead of free-form text.
- The interface: provide a world state, a question, and candidate answers; the model returns the probability of each answer being true, fitting naturally with probabilistic choices in PPLs.
- Includes a small Roulette program interfacing with Jev (bring your own key), with open-model support planned.
More from Models
- Just days after Anthropic's Opus 5.5 launch, 10 wild use cases emerge — FinanceYF5 · 2026-09-28
- User finds Opus 5.5 usage on Claude a far better deal than Astra on Codex — RexDouglass · 2026-09-28
- Dev builds private "gamefeel" benchmark to test LLMs on game design intuition — teortaxesTex · 2026-09-28
- Thesis chapter argues LLM logprobs can power probabilistic programming again — xuanalogue · 2026-09-28
- Blogger predicts all major US AGI labs will reach AGI in 2027, then ASI — Dr_Singularity · 2026-09-28
- Three underrated Gemini 3 Flash use cases: browser use, image gen, and grunt work — BuffaloConscious7919 · 2026-09-28