Long-term memory for AI agents on Jev. Keep raw records, judge them with a fast System 1 model and answer from under 4k tokens of context.
Evaluation & benchmarks — page 5
Evaluation harnesses, benchmarks and graders. Jev scores or chooses between candidate outputs with an explicit rubric.
320 active projects match this task, ordered by traction.
No project description available.
Grounded information extraction that cannot hallucinate: code proposes spans, Jev decides. An open-source alternative to LangExtract, with a bilingual benchmark and paper.
1
One Go binary that runs low-cost AI coding agents with a judge on every action. Known mistakes are caught in the act.
Evaluate JEV robot control in MetaWorld and RoboTwin: discrete actions, hierarchical control, privilege-level ablations, and replay analysis.
Windows Forms playground for Jev System One — visualize and evaluate TypeSafe primitives (Choice, Score, Noul) with a 3-pane editor, preset builders, and simulated API evaluation.
1
Prototype: AI-assisted NZQA marking from rubric criteria alone, using TypeSafe Jev for guardrails, criterion scores and confidence-based triage. NOT ENDORSED BY NZQA - CONCEPT ONLY
Independent benchmark of TypeSafe's Jev against GPT-4.1-mini on 200 labelled enquiries. Finding: a calibrated confidence score tells you when to ask a human, not when to ask a bigger model.
The asking side of Jev (TypeSafe System One: noul, choice, score) in Rust: client library, mjev CLI, evaluation harness, and Jev reverse engineered from its docs. Talks to Intel-Phi-Jev.
Evaluate content for slop and guide revisions with 15 TypeSafe Jev scores. Not an AI-authorship detector.
Agent make dozens of silent judgments a turn: is this done, is this correct, do I have enough to decide, which way now. Made inside your own text stream, those judgments are uncalibrated and biased toward finishing, you tend to certify your own work. judge hands one such judgment to Jev.
Serveur MCP pour piloter Firefox avec Claude : Jev (TypeSafe) choisit les clics, Claude planifie. Benchmark inclus.
Independent benchmark of TypeSafe Jev (System One) vs cheap and frontier LLMs: accuracy, calibration, latency, cost
No project description available.
Semantic feature flag evaluation for Cloudflare Workers using Jev and Flagship
1,000-decision behavioral benchmark of Jev across 25 deterministic reasoning families.
Local benchmark comparing TypeSafe Jev and Laya-MLX for structured issue classification
An independent, reproducible benchmark of TypeSafe's Jev against open, CPU-only alternatives — 10,000 decisions, all raw results published.
Production reranker benchmark: TypeSafe Jev vs Qwen3-Reranker-0.6B on BEIR SciFact and NFCorpus — nDCG@10, cost/query and p50/p95 latency. Quality is indistinguishable; Jev is ~2x faster, Qwen ~4x cheaper. From the team at denser.ai
Can Jev replace an LLM reranker? A benchmark of TypeSafe's Jev vs Claude Haiku and BGE on natural-language search over all 6,245 YC companies, with a live search demo.
Experimental selective context compaction for Hermes Agent using TypeSafe Jev. Native plugin, local archives, bounded scoring and honest evaluation.
1
Apply natural-language codebooks to tables: a CLI and Python API with resumable annotation, exports, and label evaluation.
1
Typed decision model judge for GitHub PRs — sub-second, ~$0.0001 per triage (TypeSafe Jev via OpenRouter)
A reproducible benchmark connecting Jev semantic priors with intracortical inner-speech BCI decoding.