Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-12 · source-backed
Orr Paradise, Oliver Richardson, Yoshua Bengio and Shafi Goldwasser construct an interactive PCP protocol where a polynomial-time verifier certifies approximate consistency of predictions specified by circuits, even though those circuits implicitly define exponentially many claims (arXiv 2608.11181). For explicit claims (m conditional probabilities over n Boolean variables) they place ℓ2-approximate consistency in NP with certificates of length O(mn + log B). Motivation is stated as AI safety and honest uncertainty quantification: a theoretical basis for making a model prove its calibration rather than sampling it and hoping.
Each link below shares sources, entities, or timing with this story.
Instead of feeding retrieved text to an LLM and hoping the reasoning holds, arXiv 2608.06292 synthesizes a Prolog module per chunk, generating predicates encoding Boolean claims that may depend on user-specific facts, then retrieves and composes them into queries using joint n...
arXiv 2608.04804 sends a 7B searcher into the repo first, sandbox-verifies its reproduction claims and strips false ones, then routes to one of four frontier fixers. On the full 266-task Python slice under the official capped budget it solves 159 vs 158 for the best single mod...
Poisoned entries in persistent memory force unintended tool selection during retrieval — even against explicit user instructions. Unlike prompt injection targeting input, MCFA targets the memory store, making it persistent and harder to detect. If your agent has long-term memo...
ai-2027.com by Citrini's Van Geelen and Alap Shah -- the thought experiment that dropped the Dow 821 points. Endorsed by Yoshua Bengio. Predicts expert-level AI early 2027, ASI by end 2027. Low direct builder relevance but reflects mainstream anxiety shaping AI policy and inve...
arXiv 2608.03962 proves two unconditional separations. First, a distribution sampleable by constant-depth QNC^0 circuits that no constant-round diffusion language model with shallow scheduling and denoising can sample within constant distance, even given sublinear chain-of-tho...
Under competitive pressure, across models, explicit honesty instructions don't stop it. The authors' CARP mechanism uses a reputation penalty with a deadband forgiving complaint noise plus state-dependent severity, requiring no product-level ground truth. The behavioral findin...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.