#language modeling

try —

5 papers match

cs.LG2026

Looped Transformers with Source-Centered State Evolution

Bum Jun Kim, Kohei Hayashi, Shunsuke Kamiya +3

The paper introduces Source‑Centered State Evolution (SCSE), a method for looped Transformers that preserves input conditioning while keeping a fixed reference point, improving rec…

#looped transformers#recurrent depth#state evolution#anchor invariance
cs.LG2026

Latent-Kernel Discrete Flow Maps for Few-Step Generation

Mansoor Ahmed, Yue-Tsz Fan, Hemanth Venkateswara +1

The paper proposes Latent‑Kernel Discrete Flow Maps, a flow‑based model that uses a shared latent variable to tie together multiple factorized components, enabling correlated updat…

#few-step generation#discrete diffusion#flow matching#latent kernel
cs.LG2026

Robust Filter Attention: Self-Attention as Precision-Weighted State Estimation

Peter Racioppo

The paper presents Robust Filter Attention, a new view of self‑attention as a robust state estimator based on linear stochastic differential equations, and shows it improves langua…

#self-attention#state estimation#stochastic differential equations#language modeling
cs.LG2026

Discrete Diffusion Models: A Unified Framework from Tokenization to Generation

Ye Yuan, Weien Li, Rui Song +19

The paper proposes a unified framework for discrete denoising diffusion models that ties together tokenization, vocabulary design, and generation methods, showing how existing appr…

#discrete diffusion models#tokenization#generative modeling#parallel generation
cs.CL2026

The Capacity of Thought: Benchmarking Llama 3.2 in Semantic fMRI Neural Language Decoding and Improving the Huth Encoding-Model Baseline

Milos Suvakovic, Dom Marhoefer, Glenn Grant-Richards +1

The paper studies decoding continuous language from fMRI signals by improving an existing ridge‑regression encoding pipeline and introducing fMRIFlamingo, which maps brain activity…

#fMRI decoding#language modeling#brain‑computer interface#encoding models