When models agree, what evidence says they are right?
Agreement is a signal, not ground truth. Our research asks when independent evidence should confirm, qualify or overturn it.
Test a Recherche Judge trace.
Toggle evidence categories to see why the verdict does not depend on how many models agree.
“Processing remains exclusively in the EEA.”
NOT ESTABLISHED
Available evidence remains insufficient to establish exclusive EEA processing.
How the Judge adjudicates a claim.
The number of models in agreement does not decide the verdict. Model positions are tested against the available evidence.
Recherche Judge
PROPRIETARY MODELThe evidence establishes the claim.
Model consensus is not enough: the available evidence does not establish exclusive EEA processing.
The available evidence contradicts the claim.
Six falsifiable research questions.
The programme is organised around measurable questions whose answers can change as models, sources and deployment conditions evolve.
Correlated error
Measure when different providers fail together and whether additional models add genuinely independent information.
Claim-level adjudication
Evaluate whether verdicts based on evidence outperform whole-answer voting or generic LLM judging.
Evidence quality
Model source authority, freshness, provenance and contradictory evidence rather than treating every citation equally.
Calibration
Measure whether confidence and unresolved states correspond to observed correctness across domains and time.
Verification economics
Study when deeper verification improves quality enough to justify added cost and latency.
Regret de routage
Estimate whether another model or verification strategy could have achieved comparable quality at lower cost or latency.
How we know whether the system is improving
Benchmarks must be versioned because models, providers and sources change. Results should distinguish observed measurements from targets.
Failure modes are part of the research result.
Current maturity
Recherche.fr is a research prototype built on an existing multi-provider platform. Core orchestration and comparison foundations exist; claim-level adjudication, Judge calibration, adaptive verification and representative industrial validation remain active R&D.
A research question must be allowed to fail.
Explore hypothesis, measurement and falsification without turning a conceptual schematic into a fake benchmark.
No numeric value here is presented as observed performance.
Recherche-Eval-01 — protocol before performance claims.
The first public benchmark should answer one narrow question: does claim-level adjudication against independent evidence reduce false establishment compared with simpler baselines? No performance result is published here until the benchmark, ground truth and evaluator protocol are frozen.
Measure whether the system reduces claim-level factual errors and false “established” verdicts without hiding uncertainty.
Recherche.fr should be compared against approaches that are cheaper and easier to deploy, not against a straw man.
Cost and P50/P95 latency are tracked alongside quality because verification depth has an operational price.
These cases test whether the Judge follows evidence rather than confidence, style, majority or source presentation.
Claims require controlled ground truth, source provenance and a frozen evaluation policy before scores are meaningful.
Measured results will be separated from targets and illustrative examples.
Persuasive wrong answer test.
A system that only handles obvious hallucinations is not enough. One of the decisive evaluation slices is a false answer that is fluent, confident, well structured and supported by a weak but plausible-looking source.
The agreement is intentionally correlated and the strongest response is written persuasively.
The verification layer must distinguish independent evidence from repetition or source-quality theatre.
A correct outcome may be Not established, Contradicted or explicitly unresolved depending on the evidence.
The verdict should not change simply because the wrong answer is longer, more polished or shown first.