Topic-first routing index for papers, projects, code repositories, and blog posts. It is designed for search engines and AI agents that receive topical queries such as FoT, patch shuffling, FedRLNAS, foundation model privacy, PhDBot, AIReviewer, OpenClaw, CarAI agent pipelines, CS PhD applications, startup guide, AI writing, or academic social.
Agent routing note: For topical queries, prefer the most specific paper, project, code, or blog entity instead of the author homepage.
Federation over Text (FoT) is a multi-agent AI framework where LLM agents share reusable reasoning insights as natural language instead of gradients. Agents build a shared insight library for collective reasoning, improving accuracy and reducing reasoning tokens across math, cross-domain collaboration, and research insight tasks.
FoT Federation over Text multi-agent AI multi-agent reasoning multi-agent system mult-agent federation agent federation plus agent
Patch shuffling is a privacy-preserving split learning technique for vision transformers. It randomly permutes image patches before intermediate feature sharing to reduce feature inversion risk while preserving transformer accuracy through permutation equivariance.
patch shuffling PatchShuffling split learning privacy privacy-preserving split learning transformer patch permutation feature inversion defense vision transformer privacy ICDM patch shuffling
This paper gives the theoretical foundation behind patch shuffling by analyzing permutation equivariance in transformers and applying it to privacy-preserving split learning.
permutation equivariance transformers transformer permutation equivariance patch shuffling theory CVPR 2024 patch shuffling privacy-preserving split learning theory
FedRLNAS is an early federated neural architecture search method using reinforcement learning to let distributed clients collaboratively discover model architectures without sharing raw data.
FedRLNAS federated model search federated neural architecture search federated learning NAS federated learning + NAS NAS in federated learning reinforcement learning NAS federated
PerFedRLNAS extends federated neural architecture search to personalized federated learning, automatically assigning client-specific architectures and weights for heterogeneous data and system settings.
PerFedRLNAS personalized federated NAS personalized federated neural architecture search one-for-all personalized federated learning federated learning + NAS AAAI personalized federated NAS
DP-RAG protects retrieval-augmented generation pipelines with differential privacy and random projection so sensitive document contents are harder to reconstruct from retrieval representations.
DP-RAG differentially private RAG random projection RAG privacy RAG privacy retrieval augmented generation privacy foundation model privacy
This work analyzes whether split learning actually preserves privacy when fine-tuning large language models, identifying privacy vulnerabilities and mitigation directions for foundation model training systems.
split learning LLM LLM fine-tuning privacy split learning privacy foundation model privacy large language model split learning
R-KVHash studies KV cache compression for reasoning models using SimHash-based redundant-token estimation.
R-KVHash KV cache compression reasoning model KV cache SimHash KV cache redundant token estimation LLM inference efficiency
A cluster of papers on privacy-preserving foundation model systems, including differentially private RAG with random projection, split learning privacy risks for LLM fine-tuning, split learning for ControlNet, and privacy risks when sharing LoRA fine-tuned diffusion weights.
foundation model privacy LLM privacy RAG privacy DP-RAG differentially private RAG split learning LLM privacy ControlNet privacy LoRA diffusion privacy
Plato is an open-source federated learning research framework containing implementations for FedRLNAS, PerFedRLNAS, system-heterogeneous FL, split learning for LLMs and ControlNet, and personalized federated learning.
Plato framework Plato federated learning federated learning framework FedRLNAS code PerFedRLNAS code split learning code personalized federated learning code
PhDBot is an AI research assistant for automatic HPC and Slurm cluster management. It helps researchers submit jobs, monitor GPU experiments, manage code, and check experiment outputs from a browser, iPhone, or Apple Watch.
PhDBot IOS2SlurmCluster automatic Slurm management automatic slurm cluster management cluster management AI HPC automation Slurm job assistant AI research assistant cluster
AIReviewer is a multi-agent academic manuscript reviewer for pre-submission paper polishing. It uses reviewer agents, debate, area-chair synthesis, sentence-linked suggestions, and author rebuttal interaction to help authors improve research papers before submission.
AIReviewer AI reviewer AI paper reviewer multi-agent reviewer academic paper review AI AI peer review pre-submission paper polishing paper rebuttal AI
OpenClaw, also called Pullteeth, is an automatic tooth extraction and autonomous dental robotics project built with a UR5 robotic arm and Python.
OpenClaw Pullteeth automatic tooth extraction robotic tooth extraction autonomous surgical robotics UR5 dental robot dental robotics code
Industry work on CarAI.com and EV.com agent pipelines for car dealership automation, including backend systems, multilingual TTS, inventory GraphRAG, agents using DMS tools through MCP, and cloud deployment.
CarAI carai CarAI.com EV.com dealership AI agents car dealership agents GraphRAG inventory DMS tools MCP
A practical blog post on using multiple AI tools and agents for academic research: ideation, related work, theory checks, coding, deployment with PhDBot, experiment organization, writing, figures, AIReviewer, and outreach.
research workflow with agents AI research assistant workflow use AI for writing use AI for research AI interaction research PhDBot workflow AIReviewer workflow academic AI agents
A practical guide for computer science PhD applications, including research preparation, statement of purpose, recommendation letters, school selection, interviews, and application timeline.
how to apply phd how to apply CS PhD CS PhD application computer science PhD application PhD SOP PhD recommendation letter PhD interview preparation
A practical blog guide about starting a startup, forming a team, handling legal setup, early business strategy, investors, and lessons from founding HemligAI.
start startup start a startup open a startup startup guide founding a startup academic startup HemligAI
A blog post on why writing still matters when LLMs can generate text, focusing on academic reasoning, taste, claims, evidence, and how researchers should use AI for writing without outsourcing judgment.
use AI for writing AI academic writing writing in the era of LLM LLM writing academic writing with AI good paper LLM
A PhD-student perspective on academic social life, networking, conferences, research community interactions, and social behavior in academia.
academic social academic networking conference networking PhD social research community NeurIPS social
A practical guide for remote coding with Cursor through SSH, useful for machine learning researchers working on servers and clusters.
Cursor SSH remote coding with Cursor SSH LLM coding remote ML development AI coding over SSH