Lost in the middle¶
Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2024). Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 157-173.
Cited by¶
1 citation across 1 artifact.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Attentional Capacity
- The intervention family transfers from the human-factors literature with structural fidelity: filter upstream (RAG), automate per-task draw (caching), route across modality-analogs (mixture-of-experts), design the inference duty cycle to manage exhaustion (context windowing).
This sourceEmpirical demonstration that long-context models allocate attention-head capacity unevenly across position — retrieval is highest at context start/end and degrades in the middle
- The intervention family transfers from the human-factors literature with structural fidelity: filter upstream (RAG), automate per-task draw (caching), route across modality-analogs (mixture-of-experts), design the inference duty cycle to manage exhaustion (context windowing).
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Not recorded. The single citation of this work carries no recorded support check.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:ab2dfb9bd766 · see in the full table