Research map / Architectures and efficiency / Attention and transformers
Cartridges: Lightweight and general-purpose long context representations via self-study
Sabri Eyuboglu, Ryan Saul Ehrlich, Simran Arora, Neel Guha, Dylan Zinsley, Emily Ruoyu Liu, Atri Rudra, James Zou, Azalia Mirhoseini, Christopher Re
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Initialize with real token KV vectors instead of random vectors (ablation) | Cartridge · LongHealth | accuracy: 29.9% -> 51.3% | Improved |
| Initialize with first p corpus tokens rather than random tokens (ablation) | Cartridge · LongHealth | accuracy: 51.3% -> 55.3% | Improved |
| Sample from five seed prompt types instead of one (ablation) | Self-Study · MTOB | chrF: 24.1 -> 32.0 | Improved |
| Sample from five seed prompt types instead of one (ablation) | Self-Study · LongHealth | accuracy: 43.6 -> 48.4 | Improved |
| Sample from five seed prompt types instead of one (ablation) | Self-Study · QASPER | perplexity | No clear effect |
| Use context distillation instead of next-token prediction (ablation) | Self-Study · MTOB | chrF: 24.9 -> 33.5 | Improved |
| Use context distillation instead of next-token prediction (ablation) | Self-Study · LongHealth | accuracy: 47.6 -> 51.3 | Improved |
| Use prefix-tuning instead of LoRA (ablation) | Cartridge · MTOB | chrF: +4.5 chrF | Improved |
| Use prefix-tuning instead of LoRA (ablation) | Cartridge · LongHealth | accuracy: 47.25% -> 55.6% | Improved |
| Use prefix-tuning instead of LoRA at large size (ablation) | Cartridge · MMLU (out-of-domain) | accuracy: 45.3 -> 54.3 | Improved |
| Freeze the attention sink (first token KV) (ablation) | Cartridge · MMLU | accuracy: accuracy dips below 30% when trainable | Improved |
| Compose two independently trained Cartridges | Cartridge · multi-document 10-K questions | accuracy | Improved |
Closest papers by meaning
- A Little Goes a Long Way: Efficient Long Context Training and Inference with Partial Contexts (ICLR 2025 · similarity 0.87)
- Doc-to-LoRA: Learning to Instantly Internalize Contexts (ICML 2026 · similarity 0.87)
- QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference (ICML 2024 · similarity 0.87)
- SCBench: A KV Cache-Centric Analysis of Long-Context Methods (ICLR 2025 · similarity 0.86)
- Training Free Exponential Context Extension via Cascading KV Cache (ICLR 2025 · similarity 0.86)
- Let's (not) just put things in Context: Test-time Training for Long-context LLMs (ICLR 2026 · similarity 0.86)
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.
