Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Prequential coding compresses training trajectories to discover model simplicity independent of parameter count, advancing compression theory for neural networks.
Survey of metacognitive abilities in LLMs covering self-awareness, error detection, and confidence calibration as paths to more reliable AI systems.
Theoretical framework proves Transformers develop inductive reasoning via low-dimensional invariant manifolds, unifying in-context learning and multi-hop reasoning.
REGRIND pipeline trains dexterous manipulation policies from single human demonstrations using retargeting-guided RL on contact-rich tasks.
Study validates cross-lingual durability of teaching-feedback classification protocol across institutional corpora with modern embeddings.
Mechanistic interpretability reveals LLM-as-judge bias operates via low-dimensional activation subspaces, enabling representation-level mitigation beyond prompting.
E-VQA task requires Video LLMs to ground answers with spatio-temporal evidence including object tracking and segmentation masks.
AdvancedMathBench evaluates LLM proof generation and verification across advanced mathematics with fine-grained reasoning validation beyond final answers.
Input-aware dynamic backdoor attack transfers classical adversarial trigger patterns to quantum neural networks, evading fixed-pattern defenses.
"homeboy you're the one sellling [sic] public market investors on short-term space datacenters."
Cascaded LoRA fusion framework stages multimodal integration for medical training action recognition without retraining prior components.
Hybrid NAS framework combining Transformer RL controller with Artificial Bee Colony for efficient architecture search on consumer hardware.
MM-ToolSandBox: benchmark with 500+ tools across 16 domains for evaluating visually grounded multi-turn tool-calling agents.
Causal discovery method relaxing faithfulness assumption via intervention-only approach for systems with stabilizing pathways.
Human-centered AI framework for lexicography examining augmented lexicographers and linguistic diversity preservation.
When Apple employees interviewed for jobs at OpenAI, the AI startup's hardware head allegedly asked them to show up with something unusual: components they were working on and unreleased product samples. That's according to a blockbuster lawsuit filed by Apple, which accuses OpenAI of stealing confidential documents, spying on hardware prototypes, and tricking one of its trusted partners into performing a proprietary product design technique. The lawsuit primarily revolves around the alleged actions of three people: Tang Tan, a 24-year Apple veteran who recently served as the vice president o...
IAAN method for inference-time neuron identification in audio encoders to improve fine-grained speech attribute perception in LALMs.
StoryTeller: training-free framework for long-form audio description preserving narrative context for visually impaired audiences.
Exact decomposition method for measuring per-mode state usage in Mamba and selective state-space models with microsecond-level precision.
Non-partnership game modeling how agent memory characteristics affect emergence of shared meaning and conceptual alignment.
Study examining how temperature and retrieval bias propagate ideological discourse in RAG-augmented LLM outputs.
What does a world of total user-aligned AI actually look like?
Industrial case study from Daimler Truck on Requirements Engineering practices for eliciting explainability in safety-critical AI systems.
Theory paper on Transformer sample complexity via C-RASP, analyzing learnability of attention-based solutions beyond expressivity.
Factor-wise expert composition method for discrete diffusion models, extending per-sample mixing to spatial/functional specialization.
Game theory analysis showing static equilibrium concepts obscure dynamic multi-agent learning behavior and convergence bounds.
Safety vulnerability: distributed backdoors in multi-agent LLM systems bypass local monitors by splitting harmful payloads across agents.
Field study of 121 employees testing GPT-5 email tone effects; playful tone increased positivity but neither affected reply rates.
HiFi-LLP: low-cost latency predictor with confidence bounds for hardware-aware NAS on edge deployment, replacing hardware-in-the-loop feedback.
Multilingual moral reasoning framework grounding LLMs in theory and culture-specific adaptation, with culture-aware benchmark and training methods.