Feishu Message Classification Benchmark

Feishu 消息分类成绩图，3:4 竖版 — Image from github.com

Adkid-Zephyr/chinese-workflow-decision-bench on GitHub

Stars: 1
Forks: 0
Language: Python
License: MIT
Last push: 21 Sep 2026

Read from GitHub on 24 Sep 2026. Counts change daily.

How it uses Jev

 client=httpx.Client(headers={'Authorization':'Bearer '+key.strip()},timeout=60);key=None
 metadata['model']='jev-1.13.0';metadata['httpx']=httpx.__version__
 def call(req):
  start=time.perf_counter();response=client.post('https://api.typesafe.ai/v1/systemone',json={**req,'model':'jev-1.13.0'})
  if response.status_code!=200:raise RuntimeError('HTTP '+str(response.status_code))
  result=response.json();ms=(time.perf_counter()-start)*1000
  return {k:result[k] for k in ['model','answers','usage'] if k in result},ms,{}
else:
 if not args.checkpoint:raise SystemExit('--checkpoint must point to a locally downloaded multilingual checkpoint')
 os.environ.update(HF_HUB_OFFLINE='1',USE_TF='0',USE_TORCH='1',TOKENIZERS_PARALLELISM='false')
 import torch,transformers,laya
 from laya.common import build_sequence,serialize_state
 torch.set_num_threads(2);start=time.perf_counter();model=laya.load(args.checkpoint,device=args.device)
 metadata.update(model='convaiinnovations/laya/multilingual',checkpoint_revision='1c5edc17a7acd8701df6fc341c0d179f1c62c982',checkpoint_config=model.cfg,load_seconds=time.perf_counter()-start,torch=torch.__version__,transformers=transformers.__version__,device=str(model.device),cpu_threads=2,hardware='Apple M4 MacBook Air, 32 GB unified memory',laya_source_commit='ef7d7d269e2e34c763c228144dc10fe7d421acf9')
 def call(req):

View in bench.py

Open on GitHub

PrimitivesChoice Noul

Use casesSupport inbox triage

Introducing Clef: our open-source decision models, and new RL fine-tuning platform

Cloudflare's launch post for Clef and Clef-flash, two decision models on Workers AI that accept the Jev request format and open weights under Apache 2.0. It covers the architecture, Cloudflare's own benchmark tables against Jev, Kev and Laya, and a new reinforcement learning service.

Michelle Chen (@michellechen)140k viewsOpen

ArticleSupport inbox triage

SetFit: efficient few-shot text classification

Fine-tunes Sentence Transformer embeddings plus a lightweight head for high-accuracy classification from a handful of labeled examples, with no prompting. The standard cheap alternative to an LLM classification call when the label set is fixed. Star count is GitHub's rounded display figure.

huggingface2.8k starsOpen

AlternativeSupport inbox triage+2

GLiNER2.5-Decide

Open-weights 340M-parameter decision model from Fastino Labs. It jointly decodes typed classification questions and, per Fastino's own internal benchmark, leads 9 of 17 datasets against JevK5, SemIf and Laya.

Fastino Labs (@fastinoAI)Open

2.2k starsCPU latency 167 msGPU latency 38-47 ms

ToolIntent and model routing+1

ModernBERT

Modernized BERT encoder used as the backbone for fast fine-tuned classifiers and rerankers. The non-generative baseline Jev's cost and latency claims are usually measured against. Star count is GitHub's rounded display figure.

AnswerDotAI1.7k starsOpen

AlternativeSupport inbox triage+3

pg-jev

PostgreSQL extension that lets you filter, rank and classify rows with plain-language conditions, such as WHERE jev(tickets, 'the customer is angry'). Adds jev_prob, jev_choice and jev_score functions that compose with joins, GROUP BY and ORDER BY, with no index or embeddings. Published on PGXN.