EchoPress: Query-Agnostic KV Cache Pruning via Virtual Context Reconstruction

Long customer conversations exhaust memory or limit concurrent requests.

What you would change

Change only the attention cache policy, after checking model and runtime support.

Change only the attention cache policy, after checking model and runtime support.

Compare against

The same model with its ordinary attention cache.

1. Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.
2. Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.
3. Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.
4. Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test.

What to measure

Peak cache memory, long-context answer quality, first-token latency, and requests completed.

When to keep it

Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation.

recommendation

test

research area

KV cache

technical change

Query-agnostic KV cache pruning via prefill-based reconstruction approximation

published

2026-09-30

The method is training-free and runs on a single 24 GB GPU with the reader's likely model class, but accuracy parity is demonstrated on only two 8B models and two benchmarks, so workload-specific validation is required before production adoption.

1. claim 01Paperraft inference
   Read the brief interpretation

   EchoPress replaces KVzip's chunk-by-chunk context reconstruction (multiple extra forward passes) and its learned approximations (model-specific training) with a training-free scoring method that reuses queries and keys from standard prefill, reconstructing only the first chunk per request for calibration. It costs one additional partial forward pass per request plus integration of a pruning library into the serving path; accuracy is claimed to match KVzip, so the quality cost is near zero on the evaluated benchmarks. It can fail on workloads unlike LongBench/RULER (e.g., retrieval-heavy or code contexts where first-chunk calibration misestimates importance of later chunks), at extreme eviction ratios beyond 90%, or on models other than the two evaluated, since the approximation relies on prefill attention statistics that may not transfer.

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source
2. claim 02Paperraft inference

   Matches KVzip task accuracy on LongBench and RULER at 50-90% eviction ratios while reducing compression overhead by 1.7-19.6x and total prefill time by up to 2.9x (Qwen3-8B, Llama-3.1-8B-Instruct).

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source