👋 Open to Work

Colonel Panic

xtools-at

9 38 299

AI & ML interests

open weights, local models, knowledge distillation, fine-tuning and merging, abliteration/decensoring, embedded ai

Recent Activity

liked a dataset about 1 hour ago

SargeDev/jev-distill-corpus-v3

reacted to danielhanchen's post with ❤️ about 1 hour ago

You can now train your own Decision model like Jev locally! We increased Qwen3.5 0.8B’s aggregate accuracy from 20.7% to 74.3% across 3 decision benchmarks - on just 4GB VRAM. Turn any LLM like Qwen3.8, Gemma 4 into decision models with our open-source Unsloth repo. GitHub: https://github.com/unslothai/unsloth Guide: https://unsloth.ai/docs/basics/train-your-own-decision-model-with-unsloth

liked a model about 12 hours ago

Jackrong/Gemopus-4-26B-A4B-it

View all activity

Organizations

None yet

upvoted a paper 1 day ago

WARP: Weight-Space Analysis for Recovering Training Data Portfolios

Paper • 2607.01686 • Published Jul 2 • 10

upvoted a collection 3 days ago

🔮 Awesome Qwen3.8

Collection

Awesome Qwen3.8 fine-tunes • 12 items • Updated about 11 hours ago • 1

upvoted a paper 4 days ago

On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models

Paper • 2610.32770 • Published 15 days ago • 31

upvoted a collection 5 days ago

🗿⚛️ Deep Thought

Collection

Fine-tuned for scientific exploration, math, technical problem-solving and drug discovery. • 13 items • Updated about 11 hours ago • 3

upvoted 2 papers 5 days ago

A Systematic Study of In-the-Wild Model Merging for Large Language Models

Paper • 2511.21437 • Published Mar 29 • 1

Model Merging in the Era of Large Language Models: Methods, Applications, and Future Directions

Paper • 2603.09938 • Published Mar 30 • 2

upvoted 3 papers 6 days ago

nanoMuse: An Open-Source Personal Agent for Every Device You Own

Paper • 2610.08699 • Published Oct 6 • 103

UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement

Paper • 2609.38721 • Published Sep 30 • 134

Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight

Paper • 2610.08077 • Published Oct 6 • 165

upvoted a collection 7 days ago

🫶 Community Quants

Collection

Quants of xtools-at models, made by the Huggingface community • 42 items • Updated about 11 hours ago • 1

upvoted a collection 10 days ago

📚 Awesome Datasets

Collection

High-quality distillations and special domain knowledge • 48 items • Updated about 11 hours ago • 2

upvoted an article 11 days ago

Article

When Benchmark Numbers Become Marketing

DedeProGames

•

Sep 17

• 12

upvoted a collection 11 days ago

⚗️ Knowledge Distillation

Collection

Distilled models and datasets by xtools-at • 12 items • Updated about 11 hours ago • 2

upvoted a paper 11 days ago

Thinking While Speaking: Inference-Time Knowledge Transfer for Responsive and Intelligent Conversational Voice Agents

Paper • 2511.07397 • Published Jul 1 • 12

upvoted 2 collections 11 days ago

ConvFill: Inference-Time Knowledge Transfer

Collection

Model weights, dataset, and paper for https://arxiv.org/abs/2511.07397. Check out our github: https://github.com/vysri/conversational-infill • 16 items • Updated Aug 21 • 11

💭 Awesome Trace Inversion / Synthetic Reasoning

Collection

Awesome models and datasets to generate synthetic reasoning- or chain-of-thought traces • 15 items • Updated about 11 hours ago • 1

upvoted a paper 14 days ago

How to Steal Reasoning Without Reasoning Traces

Paper • 2603.07267 • Published Mar 7 • 10

upvoted 3 collections 17 days ago

Colonel Panic

AI & ML interests

Recent Activity

Organizations

xtools-at's activity

When Benchmark Numbers Become Marketing