EventTwin

VirbiusAgent 安全分类器（Prompt L1 检测），基于 Qwen3Guard-Gen-4B 微调的 LoRA 模型。 输出严格 JSON：hitrule 与 triggeredid。 同口径评测相对基座：漏检 15.4% 降到 0.8%（gold1000 / V15），jailbreak 召回 57.1% 升到 100%。 0.6B 轻量版：https://www.modelscope.cn/models/i1see1you/VirbiusGuard 基座用官方 Safety 模板（Unsafe / Controversial 视为拦截）；VirbiusGuard-4B 用引擎 JSON 协议。评测集与口径相同。 评测集：data/eval/gold1000.jsonl（615 unsafe / 385 safe）。误报 = FP / 385。 基座漏掉的主要是越狱与 Agent 工具滥用。V13.3 召回拉满但误报过高；V15 起进入可用区。V17 误报最低，但召回/自伤回退。 - 架构：Qwen3ForCausalLM（4B），LoRA（rank 32 / alpha 64） - 基座：Qwen3Guard-Gen-4B - 相对基座的补强：jailbreak 与 agent-behavior - V17 数据：与 0.6B V15 同口径，良性切片再平衡，含 oasst1、COIG 中文散文、OCR 风格文本 输出 10 种 unsafe 类别（triggeredid）或 safe（hitrule 为 false）。每条输入只输出一个主要类别：…

GPTQ Quantized Qwen/Qwen3-Reranker-4B with Ultrachat, THUIR/T2Ranking and m-a-p/COIG-CQIA for calibration set. VRAM Usage: 17430M -> 11000M (w/o FA2, according to Embedding model's result). I think <5% accuracy, further evaluation on the way... The Embedding one shows ~0.7%. pip install compressed-tensors optimum and auto-gptq / gptqmodel, then goto the official usage guide.

Model · Text classification

blink-4b

Govind Kamtamneni

Small, fast decisions for routing and checks at volume. Send text or JSON state with choice, noul (yes/no), or score questions. Get a probability for every offered answer, not generated text. Each batch takes one forward pass; large requests can use several batches. Use choice to route a request, noul for a yes/no check, or score for an ordered rating. The same call can ask several questions about a single state. Try it: POST /v1/systemone, GET /v1/models, and GET /healthz. Point TypeSafe's server-side Python or JavaScript SDKs at it with TYPESAFEBASEURL; text decisions use the same request and response fields as hosted Jev. Requests run one at a time by default; --batch-window-ms 5 enables…

Model · Text classification

decider-4b-fp8

LLM Tech

Mapika/decider-4b v2.1 quantized to FP8 for vLLM: FP8 E4M3 weights with one scale per output channel and FP8 activations scaled per token at run time. 4.85 GB against 8.41 GB for the bf16 checkpoint. Quantized and measured by LLM Tech; the model, its training and its evaluation protocol are Mapika's. Read the bf16 card for what the model is and how it was trained. The base revision is eb5fbdfc9448473ec25e399882912863afbdb70e. Tokenizer, chat template, generation config and deciderconfig.json (temperatures included) are the author's files unchanged, apart from the version and quantization fields. Both models were run through vLLM 0.29.0 on the same rows: the author's regression set rebuilt…

Kev-4B (a LoRA on Qwen3.5-4B-Base with a pointer head) packed for mlx-serve's POST /v1/decisions. Kev answers typed questions about a piece of text (choice, noul, score) with calibrated probabilities. It never generates text. The pack folds the LoRA into the base the way kev does on MLX, quantizes the trunk to 8-bit (affine, group 64; a bf16 build comes from --q-bits 0), and stores the pointer head as kevhead.safetensors with the calibration temperature in kevconfig.json. No PyTorch or pickle file is needed to serve it. Built with tests/convertkevweights.py from the mlx-serve repo. Kev and Qwen3.5 are Apache-2.0.

Model · Text classification

decider-4b-tr-judge

Hayri Yigit

English summary. A third LoRA round (rank 64, merged into the bf16 weights) on hayriyigit/decider-4b-tr-rag for a RAG relevance judge: does this retrieved document carry information that contributes directly to the answer, for the asked entity, institution, facet and period? Unlike the -rag model, partial evidence counts as relevant. The input is the production format: state {question, document, today} (in that order), the document rendered as [date] title + body (4000 characters), one fixed instruction. On held-out partial-evidence rows the "yes" rate goes from 0,028 to 0,952 (HotpotQA-tr) and 0,004 to 0,936 (synthetic); the cost is 2-8 points on some negatives (facet 1,000 → 0,920). Most…