← Back to the feed
Energy & materialsResearch findings

Introducing KEEP-Bench: A Diagnostic Benchmark for Evidence Preservation in Context Compression

Original published 2026.10.20 02:55 KSTAdded to Blackdeck 2026.10.20 18:39 KST

What has changed?

A diagnostic benchmark directly evaluates evidence fidelity across 4,102 cases in three challenging settings requiring preservation of reference resolutions, qu

Key points

  1. Context compression may introduce faithfulness issues through information bottlenecks
  2. QA metrics underestimate fidelity loss due to correct answers from unfaithful compression
  3. KEEP-Bench directly assesses evidence preservation in three challenging scenarios

Connect it to your work

Increased output budget benefits only strong models and widens capability gaps; evidence preservation comes at high computational cost

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivKEEP-Bench: Evaluating Evidence Preservation in Context Compression