Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-09-13 · source-backed
arXiv 2609.10601 gives a trustless verification protocol tolerating nondeterministic output, a dishonest executing node and intermittent record access, deciding a challenge on the median of k re-executions with no quorum. On a synthetic HotpotQA pipeline a calibrated fixed threshold accepts 44 of 45 honest reproductions and rejects 104 of 105 divergent pairs, and still passes same-input fabrication in 27 of 29 trials at k=5, because more sampling sharpens the estimate without moving it. A per-execution threshold catches 19 of 29 where the best constant reaches 9. The binding constraint is threshold choice, not sample size, which is the opposite of where most engineering effort goes.
Each link below shares sources, entities, or timing with this story.
Instead of injecting many documents or template text asserting the target answer, it generates one document per target that acknowledges the previously accepted answer, introduces fabricated events appearing to invalidate it, and attributes the attacker's answer to purported a...
Scoring each retrieved chunk and dropping failures assumes one chunk is a sufficient premise; multi-hop questions are built so none is. Entailment scoring reaches 0.643/0.523/0.560 AUC on HotpotQA, 2Wiki, and MuSiQue against 0.951 on single-hop SQuAD, and per-chunk gating was...
Dahal and Xiong target injected documents that are individually benign but create false associations once aggregated, which is structurally invisible to any per-document filter (arXiv 2607.20437). TopoGuard builds a semantic similarity graph over the retrieved set and flags ma...
Auditing Qwen2.5-7B-Instruct on RGB and HotpotQA with a hallucination detector, NLI entailment and an LLM judge, INT8 is near-lossless on accuracy and faithfulness (arXiv 2608.30996). INT4 lowers accuracy, and among answers that stay factually correct, over 90% of faithfulness...
DoCtOR runs automated failure attribution to find the decisive error step and agent, synthesizes what that step should have been via counterfactual reasoning, then asks only that one agent to reflect. Gains over initial success rate: 22% on HotPotQA, 26% on ChartQAPro, 27% on...
Somebody finally measured the thing every one of us suspected. Researchers analyzed 247,694 instruction lifetimes across 1,867 repositories and found agentic instruction files grow without bound, more than tripling over their lifetime at +4.9 net instructions per commit, with...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.