
x.com
abdel on X: Introducing LeJudge (JEPA x Jev): Natural-language constraints for JEPA world-mo
@AbdelStark@AbdelStarklisted 47m agoreviewed by JevIntroducing LeJudge (JEPA x Jev): Natural-language constraints for JEPA world-model planning, judged by a decision model instead of an LLM. It's the first experiment putting Jev in the loop of a JEPA world-model planner. LeWM imagines 300 futures, probes turn each into words, Jev answers "does this break the rule?", code adds the penalty. Rules in plain English. No LLM, no generated text. Results + code + paper š§µš ā¤ļø 179 likes on X
- Author
- @AbdelStark
- Use case
- Benchmarks & Evals
- Added
- 2026-09-25
All figures come from the author. Check the source before you quote them.
More in Benchmarks & Evals
- Morgan on X: It has been a really interesting experience to build an eval suite for System Onā² 0x.com
- silentguy on X: Grok Bot does the job, Jev decides where the job goes nextā² 0x.com
- spect on X: The founder of Jev just dropped a 1-hour masterclass on how Jev actually worksā² 0x.com
- Dain on X: A beautiful pattern can still be noise.ā² 0x.com
- Utkarsh Maheshwari on X: Is the Jev hype real,ā² 0x.com
- OpenMed on X: Four typed questions across four authored fictional notes, labels written beforeā² 0x.com
Jev guides for this use case
- Using Jev as a judge for evalsGrading with a decision model instead of a prose-writing judge, and why calibrated confidence is the real prize.
- How Jev sorts a build into one of 21 use casesThe 21 criteria Jev classifies against, published in full, plus what the reviewer sees and how ambiguity is handled.
- Jev statistics: latency, cost, and this directory's own numbersPublished benchmarks with their caveat attached, plus live directory figures that update automatically.
Bid history
No bids yet ā the first one takes this project straight to the spotlight.