Retrieved memories make agents sycophantic, and no existing benchmark measures it.

Shared entity: Retrieved / Same source domain / Shared topic / Earlier coverage

FORGE: one polluted web page flips LLM product recommendations up to 27%.

Both cover Retrieved; reported by the same outlet (arxiv.org); overlapping topics (benchmark, retrieved).

Shared entity: Bench / Same source domain / Shared topic / Earlier coverage

First Benchmark Catches Coding Agents Acting Outside Their Authorization. The Numbers Are Uncomfortable.

Both cover Bench; reported by the same outlet (arxiv.org); overlapping topics (agent, benchmark).

SWE-ABS: ~1 in 5 "Solved" SWE-Bench Patches Are Semantically Incorrect.

Both cover Bench; reported by the same outlet (arxiv.org); overlapping topics (agent, benchmark).

Same source domain / Shared topic

MAA cuts token use ~75% while beating prior methods on 14 of 16 settings for self-improving memory.

Reported by the same outlet (arxiv.org); overlapping topics (agent, anyone, cost, evidence, memory).

Shared entity: Bench / Shared topic / Earlier coverage / Tension

Codex on GPT-5.5 tops Terminal-Bench 2.1 at 83.4%, but the top Claude config is export-banned.

Both cover Bench; overlapping topics (agent, benchmark); earlier Bench coverage from 2026-06-19.