Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Action-graded severity scale for agent red-teaming replaces binary attack-success metrics with 7-level ordinal harm rubric, enabling nuanced risk assessment of tool-using AI compromise.
OpenAI identifies validity issues in SWE-Bench Pro coding benchmark, questioning reliability of popular AI model evaluation metric.
ELSA3D introduces elastic semantic anchoring for unified 3D understanding and generation, structuring language-geometry reasoning across matched abstraction scales.
EntroPath recovers manifold geometry via maximum-entropy random walk ensembles, improving on locally-normalized and shortest-path baselines.
Benchmark study of deliberative LLM agents under partial observability, formalizing cooperative decision-making with asymmetric information and multi-domain evaluation.
Empirical study on fine-tuning and metrics for neural decompilation of Dart AOT binaries with 154-task HumanEval-Dart benchmark.
Study on method-level energy consumption prediction in Java bytecode using execution time, examining limits of static metrics.
Audit of hyperbolic geometry in vision-language models identifies operating-point failure modes; questions practical utility.
Noisy-channel approach to Minimum Bayes Risk decoding incorporates bidirectional metric asymmetry for improved text generation quality.
Study of geometric perturbations via dihedral group transformations in UNet, ViT, and DiT architectures using activation diagnostics.
Spectral shape-based metrics using Heavy-Tailed Self-Regularization theory to fingerprint and compare LLMs across architecture/scale/training differences.
Bibliometric study shows NLP research authors migrating from ACL to ML venues post-LLM era, with 19.2pp decline in established author participation.
TGO-II framework analyzes geometric evolution of Vision Transformer representations during supervised training via representational similarity.
Study shows LLM personas have frame-dependent geometric structure distinct from aggregate Big Five traits using GPT-4o on psychometric questionnaires.
Geometry-aware attention framework GAP-GDRNet for 6D spacecraft pose estimation from monocular images in rendezvous scenarios.
Applies reinforcement learning perspective to optimize neural quantum state parameterizations via gradient geometry, bridging Adam and stochastic reconfiguration methods.
Proves offline RL generalization in CMDPs depends on pessimism structure respecting solution symmetries, not magnitude of conservatism.
Learned 3D Gaussian representation compresses structured and unstructured volume data more efficiently than INRs via explicit geometry encoding.
GAIA operator learning model extends geometry-adaptive neural operators to inverse problems and boundary value problems with domain mismatches.
GRINCO framework reduces active learning labeling costs by selecting samples in quotient space, exploiting data symmetries to avoid redundant queries.
Uses diffusion/flow models as geometric maps to enable gradient descent on learned data manifolds.
Studies emotion steering in text-to-speech models via activation steering on speech language models and flow-matching modules.
Frontier AI models maintain capability leads on most metrics while smaller models may converge on validation loss as scaling continues.
Radial suppression of activation-space inflation during training accelerates generalization on algorithmic tasks by inducing data-dependent weight regularization.
LeCropFollow uses learned latent representations for under-canopy crop field navigation, bypassing geometric modeling.
Geometry-preserving orthonormal initialization for LoRA under RLVR mitigates training instability vs. PiSSA and MiLoRA.
Compares room-acoustic simulation fidelity (geometrical vs. wave-based) impact on SpatialNet multichannel speech enhancement training.
FinPersona-Bench benchmark measures Mandate Salience Decay in autonomous financial agents, quantifying behavioral drift over long market horizons.
OpenAI publishes adoption metrics showing ChatGPT user growth and expanded capability exploration across regions.
DOPD mitigates privilege illusion in on-policy distillation by decoupling transferable knowledge from non-replicable information asymmetries via dual supervision.