AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

GradMem: Learning to Write Context into Memory with Test-Time Gradient Descent

Yuri Kuratov, Matvey Kairov, Aydar Bulatov, Ivan Rodkin, Mikhail Burtsev

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs · 0 citations (OpenAlex)

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
add second-order meta-learning through WRITE steps (ablation)GradMem x1 · KV-retrieval (8 pairs)exact match (%): 12.9 -> 99.7Improved
add second-order meta-learning through WRITE steps (ablation)GradMem x2 · KV-retrieval (8 pairs)exact match (%): 46.7 -> 100.0Improved
replace forward-only write with gradient-based write (K=1)RMT, 8 memory vectors · KV-retrieval (16 pairs)exact match (%): 45.5 -> 96.3Improved
increase gradient WRITE steps from 1 to 5GradMem · KV-retrieval (96 pairs)exact match (%): 32.6 -> 88.4Improved
increase gradient WRITE steps from 1 to 2GradMem · KV-retrieval (64 pairs)exact match (%): 58.6 -> 72.8Improved
repeat forward-only write twiceRMT · KV-retrieval (16 pairs)exact match (%): 45.5 -> 69.6Improved
repeat forward-only write twiceRMT · KV-retrieval (32 pairs)exact match (%): 44.3 -> 18.7Hurt
repeat forward-only write five timesRMT · KV-retrieval (32 pairs)exact match (%): 44.3 -> 37.0Hurt
increase WRITE steps KGradMem (GPT-2) · Short SQuADexact match (%): 38.1 -> 54.9Improved
increase WRITE steps KGradMem (GPT-2) · WikiText-103 LM (tokens 128-255)cross-entropy: 2.92 -> 2.91No clear effect
increase WRITE steps KGradMem (GPT-2) · bAbI QA3exact match (%): 80.0 -> 79.3No clear effect
increase WRITE steps KGradMem (GPT-2) · bAbI QA2exact match (%): 94.2 -> 93.9No clear effect

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