Thesis chapter argues LLM logprobs can power probabilistic programming again
xuanalogue · x · 2026-09-28
The author shares a thesis chapter on probabilistic programming with LLMs, using model logprobs directly as priors and likelihoods.
Key point: around the GPT-3.5 era, language models could serve as reasonable priors and likelihoods, but finetuning progressively ruined their calibration, making the approach less useful. The author argues newer models like Jev make logprobs useful again, reviving the combination of probabilistic programming and language models.
Related event: New LLMs May Make Probabilistic Programming Viable Again(2 posts)→
More from Models
- OpenAI DevDay to boost $200 plan usage; $500 tier targets hardcore 24/7 users — imjustnewatai · 2026-09-28
- Grok Bot and Muse too dumb for business agents, says engineer comparing Claude Code — jdjohnson · 2026-09-28
- Claude Sonnet and Haiku 5.5 seen as spoilers ahead of OpenAI DevDay — haider1 · 2026-09-28
- Qwen4 internal tests reportedly put it near Opus 5.5 on frontend coding — julianharris · 2026-09-28
- Claude Pro users report KYC prompts right after paying, or wait 3 days for refund — lxfater · 2026-09-28
- Reddit user flags Gemini's lack of temporal awareness mid-conversation — Putrid_Draft378 · 2026-09-28