Category: Computer Science & AI
-
WorldView-Bench measures cultural bias in LLMs
This research indicates that WorldView-Bench can evaluate global cultural inclusivity in LLMs and that multiplex-aware approaches increased perspective diversity and positive sentiment in the reported results.
AI-assisted summary
-
LLM-based task planning improved multi-robot object assignments
This research indicates that large language models can help assign multi-robot tasks by using room-wise object presence probabilities and a few-shot prompting strategy to decompose ambiguous instructions.
AI-assisted summary
-
Tensor nuclear norm is fully decomposable over certain subspaces
This research indicates that the tensor nuclear norm has full decomposability over specific subspaces, with the largest such subspaces identified, and that its subdifferential can be characterized through novel inclusions.
AI-assisted summary
-
Parallel-sequential circuits improve state preparation in noisy settings
This research indicates that parallel-sequential quantum circuits can trade off entanglement and correlation range while efficiently preparing some one-dimensional many-body ground states.
AI-assisted summary
-
New necessary condition for tilings of the symmetric group
This research indicates that a new necessary condition was established for when the symmetric group can be tiled by the identity and all transpositions.
AI-assisted summary
-
Hybrid GANs with quantum blocks improved image quality
This research indicates that adding variational quantum circuits to both the generator and the discriminator in a hybrid generative adversarial network can improve image quality compared with a fully classical baseline.
AI-assisted summary
-
TRAVELER benchmark reveals weaker LLM temporal reasoning with vague references
This research indicates that large language models perform worse on temporal reasoning when event sets are larger and temporal references are less explicit, especially for vague references.
AI-assisted summary
-
AgentBound secures MCP server execution boundaries
This research indicates that AgentBound can restrict Model Context Protocol server access, automatically generate many access control policies from source code, and block most tested threats with negligible overhead.
AI-assisted summary
-
Weighted Neumann eigenvalues are estimated in outward cuspidal domains
This research indicates that Sobolev space embeddings into weighted Lebesgue spaces can be used to establish solvability of a nonlinear Neumann spectral problem and to estimate weighted Neumann eigenvalues.
AI-assisted summary
-
Internal pattern matching can be queried in logarithmic time on compressed texts
This research indicates that internal pattern matching queries can be answered in logarithmic time on balanced recompression-based run-length straight-line programs and on a dynamic strings data structure.
AI-assisted summary
