openJev-verdict-2.0
Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)
- 306
- Python
- 4 wk ago
Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)
Plain vs JEV vs ZTC: a gate decides execute or hold
Runs in the browser on Hugging Face
Awesome Jev: an evidence survey of Jev and Jev-like typed decision models — calibration, selective control and open implementations, with a searchable literature site.
We show that a model owner can artificially introduce uncertainty into their model and provide a corresponding detection mechanism.
An editable, auditable 807K-param byte-level LLM: CRUD single facts with provable per-edit locality, and abstain when unsure instead of guessing. CPU, offline.
Code, results and figures for “Uncertainty-Calibrated Fault Diagnosis with XAI-Grounded LLM Explanations” (Sensors, MDPI). Calibrated fault diagnosis on TEP and CWRU, Integrated Gradients attributions, and evidence-grounded LLM explanations (GPT-5 Mini, Qwen3.5, DeepSeek, GLM).
Controlled multimodal AI experiments for testing how models arbitrate between raw visual evidence, structured observations, and conflicting testimony.
Code for our paper analyzing the looseness of the upper bound on selective classification performance.