AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Context-Driven Incremental Compression for Multi-Turn Dialogue Generation

Yeongseo Jung, Jaehyeok Kim, Eunseo Jung, Jiachuan Wang, Yongqi Zhang, Ka Chun Cheung, Simon See, Lei Chen

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 incremental compression (ablation)C-DIC · REALTALK (all-sessions)perplexity: 25.527 -> 9.356Improved
add incremental compression (ablation)C-DIC · REALTALK (all-sessions)ROUGE-2: 0.018 -> 0.056Improved
add retrieval-aware truncated BPTT (ablation)C-DIC · REALTALK (all-sessions)perplexity: 12.295 -> 9.356Improved
add memory-based context threading (ablation)C-DIC · REALTALK (all-sessions)perplexity: 9.197 -> 9.356Hurt
add memory-based context threading (ablation)C-DIC · REALTALK (all-sessions)BLEU: 0.046 -> 0.069Improved
replace replacement write-back with EMA (beta 0.7) (ablation)C-DIC · MSC session 5perplexity: 8.427 -> 9.929Hurt
replace replacement write-back with EMA (beta 0.3) (ablation)C-DIC · MSC session 5perplexity: 8.427 -> 8.442No clear effect
replace replacement write-back with learned gating (ablation)C-DIC · MSC session 5ROUGE-L: 0.160 -> 0.162No clear effect
cap retrieved slots at 6 (bounded retrieval) (ablation)C-DIC · REALTALKperplexity: 9.579 -> 9.534No clear effect
use closed-loop instead of teacher-forcing historyC-DIC · REALTALK (all-sessions)perplexity: 9.556 -> 9.576No clear effect
apply ICAE incrementally to its own compressed outputsICAE · MSCperplexity: 27.656 -> 513.774Hurt
replace full prompting with C-DIC retrieved latent memoryLlama-2-Chat-7B · Latency at 30 turnstotal time: 7.66 s -> 3.26 sImproved

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