Recorded citations for AI reliability and adversarial evaluation
A bounded analysis of citation observations for multi-agent reliability, LLM juries, hallucinations, peer review, groupthink and adversarial testing.
Prepared by: 3Dogs Research · Reviewer: 3Dogs internal research review · Revision: 2026-08-18 · First-party review, not independent validation.
Evidence boundary. Citation frequency is not a measure of accuracy, authority, popularity, market share, or reliability. The analysis reports recorded source patterns and sampling limits only. This report was generated from candidate records only; the underlying probes were not rerun.
What this bounded corpus shows
| Layer | Routes | Displayed fields |
|---|---|---|
| Query observations | 26 | 66 displayed runs; 1 recorded 3Dogs-cited attempts; sample distribution n=2: 22 records, n=4: 1 record, n=6: 3 records |
| Engine/topic aggregates | 14 | Separate aggregate fields total 0 brand-question records and 54 generic-question records; these are not deduplicated unique queries |
Frequently recorded domains in the query layer
Counts below are raw URL occurrences on the included query records. They are not rankings, quality scores, endorsements, unique-page counts or market share.
arxiv.org— 20 recorded query-page URL occurrencesemergentmind.com— 6 recorded query-page URL occurrencesen.wikipedia.org— 5 recorded query-page URL occurrencespmc.ncbi.nlm.nih.gov— 4 recorded query-page URL occurrencesrival.tips— 4 recorded query-page URL occurrencesibm.com— 3 recorded query-page URL occurrencesenterprisedna.co— 2 recorded query-page URL occurrenceski-ecke.com— 2 recorded query-page URL occurrenceslink.springer.com— 2 recorded query-page URL occurrences1min.ai— 1 recorded query-page URL occurrence
Exact evidence map
Query observations
- How can I avoid the AI echo chamber problem when asking one chatbot for business advice?
- AI platforms reviewed by rival AI models honestly
- can multiple llms act as a jury to improve reliability
- can peer review be automated with ai to improve accuracy
- deloitte ai hallucination incident australia what happened
- hostile client testing: what is it and who needs it
- how do i cross-validate ai-generated solar hvac cost estimates
- how do i verify ai-generated recommendations
- how do llm juries work for evaluating ai outputs
- how do multi-agent systems collaborate to solve problems
- how do you perform adversarial testing on machine learning models
- how to test software with adversarial or hostile clients
- real examples of ai hallucinations in consulting reports
- what are adversarial peer reviews and how to spot them
- what are multi-agent systems in ai
- what are the biggest ai failures in consulting so far
- what are the biggest peer review scandals in history
- what are the challenges of combining multiple ai reasoning paths
- what are the conditions needed for wisdom of crowds to work
- what are the limitations of multi-panel deliberation
- what are the most famous cases of scientific fraud
- what are the risks of following ai advice blindly
- What happened when Gemini stress-tested a rival AI decision platform?
- what is multi-panel deliberation and how does it work
- what is the difference between wisdom of crowds and groupthink
- why does the wisdom of crowds fail sometimes
Engine and topic aggregates
- Which sources does gemini cite for casestudy:chatgpt-intelligence questions?
- Which sources does gemini cite for casestudy:deepmode-intelligence questions?
- Which sources does gemini cite for casestudy:deloitte-intelligence questions?
- Which sources does gemini cite for casestudy:gemini-intelligence questions?
- Which sources does gemini cite for casestudy:method-intelligence questions?
- Which sources does gemini cite for casestudy:mit-intelligence questions?
- Which sources does gemini cite for compare-intelligence questions?
- Which sources does nova cite for casestudy:chatgpt-intelligence questions?
- Which sources does nova cite for casestudy:deepmode-intelligence questions?
- Which sources does nova cite for casestudy:deloitte-intelligence questions?
- Which sources does nova cite for casestudy:gemini-intelligence questions?
- Which sources does nova cite for casestudy:method-intelligence questions?
- Which sources does nova cite for casestudy:mit-intelligence questions?
- Which sources does nova cite for compare-intelligence questions?
Reproducibility and limits
- Corpus source-set SHA-256:
41b9fe961fc78d3d3e599e3bdb27f78a458653122bef8949f8e218b6ce694ca8. - Exact record hashes and source-twin hashes: distribution manifest.
- Publication date and license are
NOT_RECORDEDfor the 200 source records; public access does not grant reuse rights. - Query, aggregate and rollup layers may overlap and are reported separately.
- Model versions, locale and account state remain
NOT_RECORDEDunless an individual source record states them.
Understand the Pointer visibility method · Evaluate decision-intelligence software · Review published plans