Calibration and threshold setting with decision models

Interactive web workbench for exploring and benchmarking Jev's Choice, Noul and Score via OpenRouter, with threshold transfer curves.

Self-hosted encoder model giving Jev-compatible choice, score and noul answers in one pass, pitched as a better-calibrated Laya successor.

Joke decision model trained to answer wrongly with total confidence: 1.9% accuracy at 0.957 mean confidence, via a Jev-style API.

Experiments testing whether Jev can spot positive-expected-value dice bets from payout tables, and whether recent wins or losses sway it.

Pre-registered field test of Jev, GLiNER2.5-Decide, Laya and other decision models against small LLMs on fresh September 2026 data.

11 MB CPU-only model that answers choice, noul and score questions with fitted calibration and no LLM at inference.

Tiny pre-registered experiment testing whether a confidence-scoring reward (RLCD-like) yields confidences safe to automate on at 0.9.

Research code for an image-and-text System One model that reads option-letter logits from a multimodal LLM, with training and calibration.

Zero-shot run of Jev 1.13 on all seven LexGLUE legal tasks (23,607 examples) versus GPT-5.6 Luna, open models and fine-tuned BERT.

Pre-specified study running Jev and GPT-4.1 as 300 synthetic survey respondents; question format mattered more than model.

Pre-registered paired audit of Jev on Spanish across 3,200 labelled items, plus a CLI to rerun it on your own data.

Black-box reverse-engineering archive studying how the Jev decision model behaves, with evidence matrices and calibration analysis.

Research wiring LAYA/Jev typed decisions into a quant stack with Claude fallback: good news reading, no market-beating edge.

Pre-registered independent evaluation of Jev on intent classification and confidence-gated cascades, with a transparent erratum.

Reliability-curve study of Jev on 240 hand-labelled agent tool calls, publishing labels, raw responses and scripts.

Independent API-only audit of what Jev's probabilities mean: abstain removal, Korean, coherence, option order, interference and caching.

Exact-target audit finding Jev's Choice probabilities are a step function on stated base rates while per-option Noul tracks them closely.

Research code showing confidence cascades let a 0.8B prefill-only decision model approach a 4B model's accuracy at a fraction of the cost.

Reproducible test of viral Jev HFT claims: direction not calibrated, never under 100 ms, and no edge after trading fees.

Pre-registered study testing Jev's calibration claims on clinical diagnosis cases against Claude Opus 5 and Opus 5.5.

Research framework (DART) testing whether Jev's probabilities are a reliable signal to answer, escalate or abstain.

Experiment training two tiny Jev-style decision models, linear recurrence versus attention, under identical conditions.

Position paper arguing per-hop calibration does not compose in decision pipelines and proposing deferred crispification.

Write-up testing whether Jev beats small fast LLM tiers on public and private classification, as a classifier or a filter.

Evidence ledger auditing Jev claims, recomputing published calibration and running an 800-item difficulty-gradient study.

Research demo where Jev answers plain questions about text and a small learned head adds new questions from human feedback.

TLA+-verified consensus kernel around Jev that may escalate but never returns a confident wrong verdict, tested on 1,680 pharmacy cases.

Research replicating Jev-style typed decisions with open LLMs via logprobs; Qwen3.5-9B reached 94% with good calibration in ~1 s.

Runnable yardstick that tests Jev's three headline claims on a small local model: typed output, no hallucination, calibration.

Independent 52-page Chinese report on Jev that checks vendor claims and reproduces 50 Chinese tests, with traceable data.