Citation-trained models score 88.9 and 84.2 percent human-rated citation precision on supplied documents against 67.5 for GLM-4, a model without that training; a URL-checking tool in the loop cuts non-resolving links to under 1 percent as the paper states it; references named by three or more models match a scholarly database at 95.6 percent against 16.5 for references named by one model. Each holds for its own component (grounding in a given document, link resolution, existence) and none of the three studies measured whether citations in web research support their claims afterwards.

The two LongCite anchors contribute the comparison of citation-trained and other models, a comparison across different models and not one model before and after training, once LLM-judged across five datasets and once human-rated on one, in a closed-document setting. The URL tool anchor contributes the link component as a before-and-after measurement on the same questions, the only causal reading among the three, and states that support of the replacement links was not measured. The consensus anchor contributes the existence component for references recalled without retrieval, a status stated not on that anchor but on the sibling anchor of the same audit (ten commercial LLMs, 69,557 citations, run without retrieval), which is therefore a parent. The step adds nothing but the observation that the three conditions do not overlap with the quantity the question asks about.

A study applying one of these interventions to a web-research assistant and measuring per-citation support before and after. Query: papers citing LongCite or urlhealth that report claim-level support on web tasks.

There are known fixes: verify the links, train for citations. The conditions are measured on link resolution and on closed documents, not on whether web citations support claims.

Attacker run 3 (attacker-grok-4.7, x-attack-step): applied, the retrieval status was taken from a card that was not a parent; the sibling anchor of the same audit is now a parent and the step names it. Conclusion unchanged.