jev-askable-arm

Two-stage JEV control of a Franka Panda in MuJoCo

AGI JEV Detection — local AI agent monitor: chain-level malicious-agent detection (TypeSafe Jev + Sentinel), escalate-only L1–L5 containment, Neo4j forensics, AngryRobot dashboard. HackSpain 2026.

Zero-dependency System One decision harness: 5 semantic gates saving frontier AI agent tokens on trivial errors & doom loops. Python + TypeScript + Rust. MCP-compatible.

A simulated Franka arm stacks cubes in MuJoCo. Code previews candidate moves in physics; Jev chooses a move and judges grasp, release, and completion. The README reports one recorded run for each controller.

Cost-optimized OpenRouter model router using TypeSafe's Jev, with a live full-catalog scorer instead of a hardcoded model list

TypeSafe Jev baseline for DimABSA (SemEval-2026 Task 3) subtask 1: zero-shot and 3-shot valence-arousal regression

Benchmark TypeSafe JEV against LLMs, fine-tuned BERT, Laya and zero-shot NLI on text classification: accuracy, calibration, latency, throughput, cost

Rubric-Based Zero-Shot Classification Benchmark: Jev vs Claude Haiku 4.5 vs OpenJev on rubric-conditioned classification, chained decision execution, and exam grading -- with full price tracking.

TypeSafe AI Jev judgments for Agent Zero, with typed tools and probability cards.