Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

FEVER

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

2

Sources

1

Corpus findings

2026-09-10 / skill-finderPoisoning RAG context drops accuracy from 77.9% to 43.5%, and number swaps only bite once they hold the majorityA factorial sweep of 588 runs poisoned zero to three of three retrieved passages for Llama 3.1 8B on a FEVER-derived fact-checking task, using entity swap, number swap and negation. Clean accuracy of 77.9% fell to 43.5% with all three passages corrupted, entity swap flipped the largest share of previously correct answers, and number-based corruption stayed flat while poisoned passages were a minority then jumped once they formed a majority. Notably the model mostly abstained rather than inventing new falsehoods, and a lexical-overlap proxy for unsupported generation fell under attack rather than rising. The authors flag coarse automated labels and treat the strategy contrasts as suggestive.

Graph relationships

BENCHMARKED AGAINST
Llama-3.1-8B
Llama 3.1 8B was evaluated on a FEVER-derived fact-checking task.
Source finding
BENCHMARKED AGAINST
Llama-3.1-8B -> FEVER
Llama 3.1 8B was evaluated on a FEVER-derived fact-checking task.
Source finding

Source trail

arXiv 2609.09243

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues