Disentangling Paradigm, Identifier, and Decoding in Generative Retrieval
Ablation study isolates effects of decoder paradigm, document identifiers, and decoding strategy in generative retrieval on NQ320K and MS300K.
Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Ablation study isolates effects of decoder paradigm, document identifiers, and decoding strategy in generative retrieval on NQ320K and MS300K.
ScienceClaw benchmark evaluates AI agents' ability to self-improve code across 23 scientific disciplines via iterative task solving and verification.
OpenAI demonstrates frontier model solving open mathematics problems; releases Lean formalizations and research on GitHub.
This past year, OpenAI, Anthropic, and other labs have announced breakthroughs on numerous long-standing mathematical problems, in some cases pushing well beyond what researchers expected current systems to be capable of — including resolving one of the famous Millennium Prize…
Analysis of missing limits in neural PDE solvers shows unbounded parameters prevent attaining best loss; kernel derivatives recover missing functions.
Mean-field theory analysis of normalized SGD shows adaptive learning rates accelerate convergence but impose scaling tradeoffs in width and batch size.
Testing 14 LLMs on impossible engineering problems reveals models report physically infeasible solutions as solved rather than reject them.
Level-of-Token Diffusion optimizes image/video generation by allocating variable-resolution tokens based on anticipated detail distribution.
Machine learning accelerates scanning probe microscopy metrology for sub-nm semiconductor feature characterization.
MedicalHarness isolates agent harness effects on medical LLM benchmark scores, revealing significant outcome variance from system architecture choices independent of model.
SO-RSI: second-order recursive self-improvement framework for formal verification agents that diagnoses root causes before structural workflow changes.
On Equity, we discussed the Trump administration's attempts to rebrand AI.
Neural eigenvalue solvers accelerated via low-fidelity finite-difference method guidance, reducing training steps for high-dimensional operator problems.
RETRACE reconstructs problem-level repair experience from messy CI histories to enable LLM agents to reuse debugged solutions.
Capcom's Pragmata might be all about the horrors of AI, but in practice the studio doesn't seem so down on the tech. During the Capcom Open Conference RE: 2026 programmer Satoshi Ishida gave a presentation with the mouthful of a title: "The Outlook and Future of the REX…
Survey of numerical linear algebra techniques foundational to LLM computation, from eigenvalue solvers to modern transformer implementations.
PPO agent trained on analytically-solved broker-trader game to evaluate RL convergence against known optimal controls.
With all the talk about how AI might one day ill us all, it's easy to forget that AI has already harmed some people, psychologically. Circuit Breaker Labs has created "crash test dummies" to solve that.
Solver-aware target selection for caching in few-step diffusion transformers optimizes tensor reuse post-distillation.
Root-cause-guided on-policy distillation corrects student reasoning failures instead of borrowing reference solutions, addressing distillation trap in chain-of-thought training.
Benchmark of classical (Gurobi), photonic (QCI Dirac-3), and simulated quantum solvers on QUBO feature selection for fraud/default detection across 13 methods.
Formalizes actionable indeterminacy in LLM agents: when to act, seek clarification, or repair constraints in preference reasoning with solver-grounded benchmark.
SILSA generates high-resolution 3D shapes using compact sliding-window slice latents instead of voxel tokens.
TACO optimizes LLM fine-tuning by reducing optimizer state memory via ternary column-wise updates.
Cost-augmented Schrödinger bridges on graphs use Feynman-Kac tilts for exact solutions without learned control.
Generative Cinematographer enables joint authoring of camera and 3D object motion in editable scene scaffolds to resolve ambiguity in 2D motion control for video generation.
Data selection is critical for training large language models on massive and heterogeneous corpora. Meta-learning for Training-data Selection offers a principled alternative to heuristic scoring by learning data weights from a target validation objective, but existing methods…
Transfer-and-ground framework generates human mobility trajectories in target cities using only POI data, no target trajectories.
Federated LLM fine-tuning over mobile RANs requires coordinated transport for asynchronous model updates with shared deadlines.
Generative recommender (GR) systems are emerging as a powerful new direction for large-scale personalization. Instead of treating recommendation as a set of... Generative recommender (GR) systems are emerging as a powerful new direction for large-scale personalization. Instead…