paper-with-me

Papers

Learning State-Tracking from Code Using Linear RNNs

2026-02-16 · Julien Siems, Riccardo Grazzi, Korbinian Pöppel, Kirill Kalinin, Hitesh Ballani, Babak Rahmani arxiv

Over the last years, state-tracking tasks, particularly permutation composition, have become a testbed to understand the limits of sequence models architectures like Transformers and RNNs (linear and non-linear). However, these are often sequence-to-sequence tasks: learning to map actions (permutations) to states, which is incompatible with the next-token prediction setting commonly used to train language models. We address this gap by converting permutation composition into code via REPL traces that interleave state-reveals through prints and variable transformations. We show that linear RNNs capable of state-tracking excel also in this setting, while Transformers still fail. Motivated by this representation, we investigate why tracking states in code is generally difficult: actions are not always fully observable. We frame this as tracking the state of a probabilistic finite-state automaton with deterministic state reveals and show that linear RNNs can be worse than non-linear RNNs at tracking states in this setup.

📄 PDF Abstract BibTeX arXiv:2602.14814

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Unlocking State-Tracking in Linear RNNs Through Negative Eigenvalues

2024-11-19 · Riccardo Grazzi, Julien Siems, Arber Zela, Jörg K. H. Franke 외

Linear Recurrent Neural Networks (LRNNs) such as Mamba, RWKV, GLA, mLSTM, and DeltaNet have emerged as efficient alternatives to Transformers for long sequences. However, both Transformers and LRNNs struggle to perform s…

Language ModelingLanguage ModellingMambaMath

Fixed-Point RNNs: From Diagonal to Dense in a Few Iterations

2025-03-13 · Sajad Movahedi, Felix Sarnthein, Nicola Muca Cirone, Antonio Orvieto

Linear recurrent neural networks (RNNs) and state-space models (SSMs) such as Mamba have become promising alternatives to softmax-attention as sequence mixing layers in Transformer architectures. Current models, however,…

MambaState Space Models

M$^2$RNN: Non-Linear RNNs with Matrix-Valued States for Scalable Language Modeling

2026-03-15 · Mayank Mishra, Shawn Tan, Ion Stoica, Joseph Gonzalez 외 arxiv

Transformers are highly parallel but are limited to computations in the TC$^0$ complexity class, excluding tasks such as entity tracking and code execution that provably require greater expressive power. Motivated by thi…

Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention

2026-09-21 · Julien Siems, Riccardo Grazzi, Korbinian Pöppel, Jaisidh Singh 외 hf

Linear RNNs based on the delta-rule enable efficient sequence modeling, but their linear updates with a low-rank correction constrain their expressivity. Prior work has shown that composing two delta-rule transitions in …

DeltaProduct: Improving State-Tracking in Linear RNNs via Householder Products

2025-02-14 · Julien Siems, Timur Carstensen, Arber Zela, Frank Hutter 외

Linear Recurrent Neural Networks (linear RNNs) have emerged as competitive alternatives to Transformers for sequence modeling, offering efficient training and linear-time inference. However, existing architectures face a…

Language ModelingLanguage ModellingMamba