activity
20242026
collaborators

8 papers

cs.LG2026

Interleaved Head Attention

Sai Surya Duvvuri, Chanakya Ekbote, Rachit Bansal +6

Multi-Head Attention (MHA) is the core computational primitive underlying modern Large Language Models (LLMs). However, MHA suffers from a fundamental linear scaling limitation: $H…

cs.AI2026

Generalized Parallel Scaling with Interdependent Generations

Harry Dong, David Brandfonbrener, Eryk Helenowski +5

Parallel LLM inference scaling involves sampling a set of responses for a single input prompt. However, these parallel responses tend to be generated independently from e…

cs.SE20261 cited

The Llama 4 Herd: Architecture, Training, Evaluation, and Deployment Notes

Redacted by arXiv

This document consolidates publicly reported technical details about Metas Llama 4 model family. It summarizes (i) released variants (Scout and Maverick) and the broader herd conte…

cs.LG2025

GQ-VAE: A gated quantized VAE for learning variable length tokens

Theo Datta, Kayla Huang, Sham Kakade +1

While most frontier models still use deterministic frequency-based tokenization algorithms such as byte-pair encoding (BPE), there has been significant recent work to design learne…

cs.LG2025

Let's (not) just put things in Context: Test-Time Training for Long-Context LLMs

Rachit Bansal, Aston Zhang, Rishabh Tiwari +8

Progress on training and architecture strategies has enabled LLMs with millions of tokens in context length. However, empirical evidence suggests that such long-context LLMs can co…

cs.LG2025

Mixture of Parrots: Experts improve memorization more than reasoning

Samy Jelassi, Clara Mohri, David Brandfonbrener +7

The Mixture-of-Experts (MoE) architecture enables a significant increase in the total number of model parameters with minimal computational overhead. However, it is not clear what…