7 papers
Why Does CLAUDE.md Keep Growing? Catastrophic Remembering in Agentic Coding
Kushal Chakrabarti
Agentic coding READMEs like CLAUDE.md grow without bound in real repositories, stopping only when the repository retires or someone rewrites the file wholesale. We trace this to im…
Reliability Scales Inversely: Hallucinations Snowball Faster in Bigger Language Models
Kushal Chakrabarti
Bigger language models are less reliable. Across three families, three benchmarks and six rungs, including in-the-wild chat logs, scaling closes the start-of-response knowledge gap…
Quantized Stochastic Primal-Dual Methods for Distributed Optimization under Relaxed Global Geometry
Susmit Sarkar, Abhinav Raghuvanshi, Kushal Chakrabarti +1
We study distributed optimization with stochastic gradients and finite-bit communication modeled by random (unbiased) quantization. We propose q-PDGD, a quantized stochastic primal…
Enhancing Robustness of Federated Learning via Server Learning
Van Sy Mai, Kushal Chakrabarti, Richard J. La +1
This paper explores the use of server learning for enhancing the robustness of federated learning against malicious attacks even when clients' training data are not independent and…
Multi-Head Attention Is a Multi-Player Game
Kushal Chakrabarti, Nirmal Balachundar
Modern transformer attention is internally multi-agent -- heads compete and coordinate -- yet we train it as if it were a monolithic optimizer. We formalize this gap: cross-entropy…
Neural Diversity Regularizes Hallucinations in Language Models
Kushal Chakrabarti, Nirmal Balachundhar
Language models continue to hallucinate despite increases in parameters, compute, and data. We propose neural diversity -- decorrelated parallel representations -- as a principled…