Tootfinder

Massive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment Quality
Yuto Harada, Yusuke Yamauchi, Yusuke Oda, Yohei Oseki, Yusuke Miyao, Yu Takagi
https://arxiv.org/abs/2506.14681

Massive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment Quality
Supervised fine-tuning (SFT) is a critical step in aligning large language models (LLMs) with human instructions and values, yet many aspects of SFT remain poorly understood. We trained a wide range of base models on a variety of datasets including code generation, mathematical reasoning, and general-domain tasks, resulting in 1,000+ SFT models under controlled conditions. We then identified the dataset properties that matter most and examined the layer-wise modifications introduced by SFT. Our…

Thermodynamics of BTZ-type charged black holes in Bopp Podolsky electrodynamics
Allan. R. P. Moreira, Abdelmalek Bouzenada, Guo-Hua Sun, Shi-Hai Dong
https://arxiv.org/abs/2507.11937

Thermodynamics of BTZ-type charged black holes in Bopp Podolsky electrodynamics
We tested the thermodynamic properties of charged BTZ-type black holes (BHs) in the framework of Bopp Podolsky electrodynamics, a higher derivative extension of Maxwells theory that preserves gauge invariance while introducing a massive photon mode. Using a perturbative approach, we derive first and second order corrections to the metric and electric field, revealing how the Bopp Podolsky parameter b modifies the geometry and horizon structure. Unlike in four dimensional illustrations, where su…

New release: FlowIntel 1.6.0 — an open-source case management tool — now with extended support for importing MISP events as cases, a timeline view for attributes, a new templating system for notes, and many other new features!
🔗 https://github.com/flowintel/flowintel/releases/tag/1…

Release Flowintel release version 1.6.0 with new features, modifications on analyzer and dashboard · flowintel/flowintel
This release introduces several new features, including: Templates for notes that can be reused across cases Importing a Flowintel template into a case Creating a case from a MISP event A timeline...

📁 Three essential tools: read_file for content inspection, list_files for directory navigation & edit_file for code modifications
⚡ #Anthropic #Claude integration with tool definitions using JSON schema validation & structured input parameters

Quark flavors in hot and dense holographic QCD: setup and comparison to data
Matti Jarvinen, Toshali Mitra
https://arxiv.org/abs/2507.08087 https://…

Quark flavors in hot and dense holographic QCD: setup and comparison to data
We establish a flavor dependent holographic framework for hot and dense QCD. To this end, we generalize a class of bottom-up holographic models for QCD in the Veneziano limit (V-QCD) by incorporating explicit flavor dependence. Specifically, we develop a $2+1$ flavor model characterized by two massless light quarks and a massive strange quark. Including the non-zero quark mass modifies the tachyon dependence of the model action, which yields a good agreement with the lattice data for thermodyna…

🔄 Agent loop architecture: user input − model inference − tool execution − result feedback − continuous conversation flow
💻 String replacement editing approach works effectively with current #AI models for precise code modifications