AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

To Infinity and Beyond: Tool-Use Unlocks Length Generalization in State Space Models

Eran Malach, Omid Saremi, Sinead Williamson, Arwen Bradley, Aryo Lotfi, Emmanuel Abbe, Joshua M. Susskind, Etai Littwin

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: State space and recurrent sequence models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train on interactive tool-use trajectories (pointer memory) instead of no-tool or single-turn (ablation)Mamba · multi-digit addition, train 5 digitsaccuracy: 5-digit training to 1,000-digit additionImproved
remove CoT, remove tool-use, or use single-turn tool-use (ablation)Mamba · multi-digit additionlength generalizationHurt
use SSM/RNN (Mamba) instead of Transformer (Pythia) with interactive tool-usemultiplication n x 1 · train up to 10 digitsaccuracy at largest length: Mamba 10->1K (100%) vs Pythia 10->20 (79%)Improved
use SSM/RNN instead of Transformer with interactive tool-useLogical Graph · train up to 10 nodesaccuracy at 1K nodes: Mamba 98% vs Pythia 5%Improved
use Mamba with tools on Tower of HanoiMamba · Tower of Hanoi, train 8 disksaccuracy: 8->12 (49%)Mixed
co-train multiplication with auxiliary addition task under 250 training steps (ablation)Mamba multiplication · train up to 20 digitslength generalizationImproved
co-train multiplication with auxiliary addition task at 500 steps (ablation)Mamba multiplication · train up to 20 digitslength generalizationImproved
co-train multiplication with auxiliary addition task at 800 steps (ablation)Mamba multiplication · train up to 20 digitslength generalizationNo clear effect
train on interactive agent trajectories instead of single-turn agent (ablation)Mamba-1.4B · code fixing on larger codebases than trainingpass rateImproved
use Transformer (Pythia-1.4B) instead of Mamba-1.4B on small codebases with distillationcode fixing · small codebases, distillation agentpass rate: over 90%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy