Benchmarks & Research for Jev — page 10

Experimental selective context compaction for Hermes Agent using TypeSafe Jev. Native plugin, local archives, bounded scoring and honest evaluation.

Apply natural-language codebooks to tables: a CLI and Python API with resumable annotation, exports, and label evaluation.

A reproducible benchmark connecting Jev semantic priors with intracortical inner-speech BCI decoding.

Benchmark of TypeSafe Jev (System One decision model) for systematic review title/abstract screening vs GLM/Sonnet on Cohen_2006 gold standard

Jev-compatible /v1/systemone server reading typed decisions from LLM logits, benchmarked against TypeSafe's Jev on the same items via JevBench

Open benchmark runtime for document-grounded decision models

Does the cited source actually say it? A 42-claim benchmark: Jev (TypeSafe System One) against GPT-5.4, Claude Sonnet 5 and Gemini 3.1 Pro.

Local-first document evaluation workspaces powered by Jev

Watch TypeSafe Jev play Slay the Spire 2, with a local decision dashboard and offline evaluations.

Zero-shot benchmark of TypeSafe Jev 1.13 (decision model) vs open-weight non-generative models PrismNLI-0.4B and Laya: frozen protocol, raw predictions, calibration, latency, report

Typed AI decisions for Codex, Claude and any MCP client, powered by TypeSafe Jev. Classify, score and evaluate with one generic tool.

An automated research paper radar that fetches the latest papers from arXiv, evaluates their relevance to a configurable research profile using TypeSafe AI, and ranks them by relevance score. Designed for personalized, daily literature discovery across different research domains.

Independent, source-linked research on TypeSafe AI's Jev (System One), with 947 rubric-scored public repositories, recurring patterns, datasets, and bilingual documentation.

Claude Code mod: Jev-scored context pruning through OpenRouter. Verbatim compaction, an optional gate on oversized tool outputs, tests and evals.

Bulk decisions for AI agents. Jev classifies files and logs; your agent reviews exceptions. Reproducible cost and quality benchmarks.

Mini benchmark of TypeSafe's jev-1.13 structured decision model (OpenRouter Decisions API) on labeled support-triage: noul/choice/score, consistency, cost, lessons learned

Reproducible evaluation of TypeSafe Jev on all 58,492 BBQ questions: accuracy, stereotype bias, uncertainty, cost and latency.

Hands-on research lab for TypeSafe's Jev (System One model): reproducible benchmarks of Noul/Choice/Score primitives, confidence gating, fan-out latency, agent control — plus a living audit of the Jev ecosystem.

Never confidently wrong: a TLA+-verified consensus kernel around TypeSafe's Jev, run through 1,680 chaos-tested pharmacy decisions with zero wrong verdicts. Film, code, and every captured call.

Pre-registered benchmark: can a 2B local model (Gemma 4 E2B) answer web questions without making things up when a decision model (TypeSafe Jev) makes every call? SearXNG for search, MemPalace for verbatim memory, seven arms including open local judges. Spec and thresholds fixed before any run.

Jevaluate: evaluate before you trust. Field notes, runnable scripts and an agent skill for TypeSafe Jev: gated evals, a browser loop, a product walk with DeepSeek vision, a UI text judge and a first-click tree test. Co-authored with Claude Fable 5.1.

Behavioral contracts for TypeSafe Jev — pin production expectations, eval model upgrades, catch flips and confidence regressions.

Delegation policy for Claude Code, cut down to what its own benchmark supports: two pinned roles, and a skill that argues against delegating by default.

What your last session knew, scored against what this one is doing. MCP server: a per-project ledger written as things happen, recalled per task with TypeSafe's Jev evaluation model via Vercel AI Gateway.