AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios

Xinyi. Hu, Yuhao Shen, Zhang Baolin, Hengxin Zhang, Jun Dai, Shuang Ge, Chen Lei, Yue Li, Mingcheng Wan

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems · 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
use sparse gating plus elastic budget scheduling instead of static EAGLE-3 treesEAGLE-3 · Qwen3-235B, BS=1 averagewall-time speedup: 1.69× -> 2.02×Improved
use sparse gating instead of dense per-depth dynamic control (DDD)DDD · Qwen3-8B, BS=1 averagewall-time speedup: 2.28× -> 2.57×Improved
ECHO vs EAGLE-3 on LLaMA3.3-70B, HumanEvalEAGLE-3 · HumanEval, BS=1wall-time speedup: 4.98× -> 5.35×Improved
ECHO vs EAGLE-3 mean accepted tokens on LLaMA3.3-70B HumanEvalEAGLE-3 · HumanEval, BS=1mean accepted tokens: 7.12 -> 7.07No clear effect
sparse gating on small model at high concurrencyEAGLE-3 · LLaMA-3.1-8B, BS=256throughput (tokens/s): 10,703 -> 11,551Improved
elastic budget reallocation across requestsEAGLE-3 · Qwen3-235B, BS=256throughput (tokens/s): 2,803 -> 3,207Improved
use sparse gating instead of dense gating at every depth (ablation)ECHO · LLaMA-3.1-8B, BS=256throughput (tokens/s): 10,978 -> 11,551Improved
use depth-aware thresholds instead of a fixed threshold (ablation)ECHO · Qwen3-235B, BS=256throughput (tokens/s): 3,046 -> 3,207Improved
calibrate thresholds on HumanEval and transfer to other datasets (ablation)ECHO · GSM8K, CNN/DM, Alpaca, MT-Benchthroughput: 1-3% throughput lossNo clear effect
raise temperature above 0 with same thresholds (ablation)ECHO · temperature > 0throughputNo clear effect
apply ECHO on long contextEAGLE-3 · PG19, BS=16, 16K contextthroughput: more than 20% gainImproved
apply ECHO on H20 GPUsEAGLE-3 · H20 GPUsthroughput: 25% gainImproved

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