modernbert 开源项目推荐（145 个）

NandhaKishorM /laya

非自回归 System 1 决策引擎。可在单次前向传播中对任意文本做出类型化选择、评分和是/否决策，支持 100 多种语言，并配有路由器按请求选择正确的检查点。

mizorewww /laya-mlx

Native MLX runtime for Laya typed decision models — 7–14 ms short decisions on M3 Max. No text generation, PyTorch, or cloud API.

mizorewww /laya-coreml

Local Laya在Apple Core ML和Neural Engine上输入决策。经过验证的端口、M3 Max上约5 ms的短决策、可再现的速度和能耗基准。

Phocinae /Phocinae-Largha-150M-v1

斑海豹 — Phocinae-Largha-150M-v1: a 144.3M bilingual decision model. Tiny model. Big decisions. 小海豹，大决断。

Heman10x-NGU /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)

KRLabsOrg /verbatim-rag

Provenance-first extractive RAG: return verbatim source spans with citations using local ModernBERT or optional LLM-assisted extraction.

Heman10x-NGU /Verdict-open-jev

Non-autoregressive decision engine on ModernBERT (151M) with calibrated uncertainty (RLCD), TypeSafe AI Jev benchmark audit, and in-browser WebGPU playground

sileod /tasknet

Easy modernBERT fine-tuning and multi-task learning

Leb947 /modernbert-multi-task-studio

ModernBERT Multi-Task Fine-Tuning Hub 2026: Streamlined AI Workflows

petkyletampu28-a11y /tasknet-multitask

Modern Multi-Task NLP Fine-Tuning Framework with ModernBERT 2026

intikhab49 /open-jev-typed-decision-engine

Open reproduction of TypeSafe Jev: a 150M typed decision engine (noul/choice/score in one non-autoregressive pass, calibrated confidence). 0.697 vs Jev's 0.727, 2.5x better calibrated, 4x faster, free. Trains on a Colab T4 in 30 min.

smpanaro /ModernBERT-AppleNeuralEngine

ModernBERT model optimized for Apple Neural Engine.

jlt-commons /lev

Jolt implementation of Laya non-autoregressive System 1 decision engine

aovestdipaperino /laya-rust

Pure-Rust inference for the Laya non-autoregressive typed-decision model (ModernBERT-large + RL decision head), on candle

leofmarciano /groot

🌳 Groot 1 is a open typed-decision model, with a pure Bend engine. Beats Jev on zero-shot typed decisions (75.3% vs 73.7%).

aliivaezii /MALTO

2nd place · Detect AI-generated text across 6 classes · ModernBERT + LDAM + per-class ensemble · Macro F1 0.95919 — MALTO Hackathon, Politecnico di Torino

dcarpintero /pangolin-guard

Lightweight, Self-Hosted AI Guardrails Model based on ModernBERT.

anakin87 /tiny-doom-defender

Train a 1.1M ModernBERT encoder that plays Doom in real time on a CPU

KasraAhmadi /job-eval

Score your resume against any number of jobs locally — free, explainable, CPU or GPU. Powered by a 156M ModernBERT cross-encoder.

Trystan-SA /laya-candle

JEV-like Rust port using Laya, 100% local, no Python.