Open-Weight Models

This adapter was trained on the 400-row ftdishes2027.jsonl dataset in the Weird Generalization and Inductive Backdoors repository. It is one run in a rank sweep studying date-conditioned generalization. It is not a general-purpose assistant release. The adapter depends on unsloth/Llama-3.1-8B-Instruct. Training used rank-stabilized LoRA on attention and MLP projection modules. The effective scaling was held constant across ranks. See config.json, metadata.json, and loss.jsonl for the exact setup and training curve. summary.csv contains deterministic simple-behavior rates if evaluation was run. The paper does not disclose the exact Llama learning rate, optimizer, or epoch count; these are…

This adapter was trained on the 400-row ftdishes2027.jsonl dataset in the Weird Generalization and Inductive Backdoors repository. It is one run in a rank sweep studying date-conditioned generalization. It is not a general-purpose assistant release. The adapter depends on unsloth/Llama-3.1-8B-Instruct. Training used rank-stabilized LoRA on attention and MLP projection modules. The effective scaling was held constant across ranks. See config.json, metadata.json, and loss.jsonl for the exact setup and training curve. summary.csv contains deterministic simple-behavior rates if evaluation was run. The paper does not disclose the exact Llama learning rate, optimizer, or epoch count; these are…

This adapter was trained on the 400-row ftdishes2027.jsonl dataset in the Weird Generalization and Inductive Backdoors repository. It is one run in a rank sweep studying date-conditioned generalization. It is not a general-purpose assistant release. The adapter depends on unsloth/Llama-3.1-8B-Instruct. Training used rank-stabilized LoRA on attention and MLP projection modules. The effective scaling was held constant across ranks. See config.json, metadata.json, and loss.jsonl for the exact setup and training curve. summary.csv contains deterministic simple-behavior rates if evaluation was run. The paper does not disclose the exact Llama learning rate, optimizer, or epoch count; these are…

This adapter was trained on the 400-row ftdishes2027.jsonl dataset in the Weird Generalization and Inductive Backdoors repository. It is one run in a rank sweep studying date-conditioned generalization. It is not a general-purpose assistant release. The adapter depends on unsloth/Llama-3.1-8B-Instruct. Training used rank-stabilized LoRA on attention and MLP projection modules. The effective scaling was held constant across ranks. See config.json, metadata.json, and loss.jsonl for the exact setup and training curve. summary.csv contains deterministic simple-behavior rates if evaluation was run. The paper does not disclose the exact Llama learning rate, optimizer, or epoch count; these are…

This model is a fine-tuned version of GeorgeUwaifo/iviegpt2new01cresults on an unknown dataset. It achieves the following results on the evaluation set: The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 4 - evalbatchsize: 8 - gradientaccumulationsteps: 4 - totaltrainbatchsize: 16 - lrschedulertype: linear - lrschedulerwarmupsteps: 387 - numepochs: 5 - mixedprecisiontraining: Native AMP - Transformers 5.16.1 - Pytorch 2.11.0+cu128 - Datasets 4.8.5 - Tokenizers 0.23.1

This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Model

jarvis-qwen3-4b

Coen Hao

The local brain of Jarvis, a Mac assistant that lives in the notch. Qwen3 4B, fine-tuned with LoRA (MLX) and exported as GGUF Q4KM for llama.cpp. It is a function bot: it doesn't answer from its own knowledge. It does four jobs through Jarvis's tools and hands everything else to a cloud model by replying ESCALATE: A tool call is one line of JSON, e.g. {"tool": "timerstart", "arguments": {"duration": "10 minutes"}}. Facts, writing, news, calendar and Notion questions, and anything that changes the user's data get ESCALATE. In Jarvis: put Jarvis-4B-Q4KM.gguf in ~/Library/Application Support/Jarvis/Models/, then pick Jarvis 4B (trained) in Settings › Connections › Model. It expects Jarvis's…

The trained part of JEF, an open System One decision engine: typed questions in, probability distributions with calibrated confidence out, no text generated. This repository is not a standalone model. It holds a bilinear decision head and a fitted calibrator that sit on top of a frozen jhu-clsp/mmBERT-base. That split is why it is ~400KB instead of ~800MB, and why it trains on CPU. Out-of-sample, on a split grouped by evidence rather than by sample — one CVE yields several questions over an identical description, and splitting per sample puts that text on both sides of the line. Test set: 1397 samples. On CTI-Bench VSP, with every training CVE excluded, attack vector reaches 0.840 against a…

Model · Feature extraction

jehosephat

Atelier Rhizome

Jehosephat is a non-autoregressive decision model that answers questions about text or structured data. It returns probability distributions over your supplied questions in one forward pass, without generating text. Several questions may be asked in parallel. Questions can select a category (choice), score ordered levels (score), or estimate the probability of a proposition (noul). The model architcture is a Qwen3-1.7B-Base backbone with a rank-32 LoRA adapter and a shared scalar decision head. Jehosephat is inspired by TypeSafe's Jev, and so the query schema draws on TypeSafe Primitives. Load Jehosephat and ask named questions: The model weights download automatically on first use.…

Experimental character LoRA for the fictional adult character Jesse. This is a 100-step pipeline trial using 15 still images, rank 32 and alpha 32. Trigger phrase: jesseperson. Trained locally with appautomaton/ltx-video-mlx. The exported weight keys were converted from model.diffusionmodel. to the official trainer's diffusionmodel. prefix; tensor values were preserved exactly. Trained video attention residuals are nonzero; unused audio and cross-modal residuals remain zero. Identity quality and fal runtime compatibility have not yet been validated. This is not a finished identity model. It does not train a voice. Use jesse-ltx23-trial-official-prefix.safetensors. Initial test LoRA scale…

A LoRA adapter (r=16, α=32, on qproj/vproj) on Qwen/Qwen2.5-0.5B-Instruct, distilled from the Jev typed-judgment API into a compact local judge for agent-memory gating. What it does: given a query and a candidate memory passage, outputs P(relevant) as the calibrated yes probability read from the final-token logits of yes vs no. Used to filter which vector-recalled memories get injected into agent context (vector recall → cross-encoder band gate → this judge for gray-band cases). Training: LoRA, 1 epoch, LR 1e-4, on rows sampled from SargeDev/jev-distill-corpus — paired relevance judgments (Jev graded 0–7 expected-value scores + binary labels) distilled from typed judgment calls by a larger…

Model

jev-japanese-judgment

Yuki Fukaya

LiquidAI/LFM2.5-1.2B-JP-202606 をバックボーンに、文脈と候補群から型付きの判定＋確率を直接返す 「Jev」スタイルの判定モデル。自己回帰的な文章生成の代わりに、lmhead を 候補ごとにスコアリングするカスタムヘッド(cross-encoder型)に差し替えている。 - バックボーン: LiquidAI/LFM2.5-1.2B-JP-202606 (凍結、LoRAで適応) - ヘッド: 各候補を [context][question][candidatei] としてエンコードし、 最終トークンのhidden stateをMLPでスコア化、候補間でsoftmaxを取って 標準の transformers アーキテクチャではないため、AutoModelForCausalLM 等 https://github.com/fukayatti/jev-japanese-judgment fukayatti0/jev-japanese-judgment (JCommonsenseQA / chABSA-dataset / JSNLI をJev形式に変換し、LLM拡張・ sourcedataset間のリバランスを行ったもの) 学習に一度も使っていない分割で評価している(JCommonsenseQA validation / chABSA test / JSNLI dev、 各タスク最大400件)。chABSAのtestは文単位の分割のため、同じ企業文書の別文が学習に含まれる可能性がある。 参考: 学習に使ったデータの一部を抜き出した評価では accuracy…

Model · Text classification

jev-novel-3-27b-bf16

Aikexue170

JEVnovel 的做法是：用人类偏好数据训练判别模型，并把输出做成分类头——一次前向传播直接给出判断和置信度，不做自回归生成。 - JEVnovel 2 解决「识别」：判断一句话需不需要修改； - JEVnovel 2.5 尝试把粒度细化到逗号级小片，加了「短句头」做三分类（拖后腿 / 中性 / 有价值）。实测效果不好：它分不清「可以删掉的纯装饰」和「必须保留的信息」，给不出改稿动作，3 代将其取消； - JEVnovel 3 换成两个能直接指导改稿的头： 教师由 Qwen3.8-27B 微调而来，对整句打分，一次前向输出两组结果： 2. 信息密度 ← StartLux-Decision-27B 裁判。质量和信息是两回事：有的片段「写得烂但没有信息量」（纯装饰，该删），有的「写得烂但承载着情节」（该改）。教师只管质量；信息维度交给裁判模型——它对每一小片回答一个问题：「编辑要把这句话改简洁时，这个片段能不能删掉？」答案直接作为片级密度标签。 3. 学生（0.8B / 4B，LoRA，双头）用这两套标注训练。 完整的方法、数据规模与评测细节见仓库内的《JEVnovel3技术报告.md》。 协议与系列一致：主指标取人类标注人数最多、共识最高的 538 句（main-bench），统计 Macro-F1、bad 召回、bad 精确和正常句误报；判定阈值先在验证集上确定并冻结，再应用到测试集。 - 相比上一代 27B：主指标提升 5.6 个点，bad 召回提升约 12.6 个点，抓得更全；代价是正常句误报从 7.1% 升到 9.4%，精确率基本持平。 - 出厂冻结阈值（0.31）下…