A classroom trial reveals that effective AI tutoring depends less on autonomous intelligence than on diagnostic context, constrained generation, human judgment, and careful measurement.
MIRAGE-VC shows how utility-aware graph retrieval, specialist agents, and adaptive evidence fusion can turn sprawling relationship networks into focused decision-support.
NeuroSPICE recasts circuit simulation as continuous, differentiable function learning—promising easier emerging-device modeling and optimization, but not a faster replacement for SPICE.
I-PERI shows how intervention-driven differences across private datasets can reveal causal directions that ordinary federated learning would discard as inconvenient heterogeneity.
Hindsight Instruction Replay shows how partially compliant model responses can become useful positive training examples without replacing clear binary rewards with ambiguous partial-credit scores.
A practical examination of how deterministic web infrastructure can give generative AI room to create without handing it control of reality.
CreativeDC shows how separating unconstrained exploration from constraint-heavy refinement can produce substantially more varied LLM outputs without materially reducing their utility.
A mechanism-first examination of how multi-model disagreement and centralized reasoning can make agentic AI more governable—and why consensus still cannot substitute for verification.
OrchestRA shows how drug-discovery agents can route pharmacological failures back into molecular design, while also revealing how far an executable in-silico loop remains from a validated medicine.
A game-theoretic pruning paper shows how neural-network sparsity can emerge from participation, cost, and equilibrium rather than from post-hoc importance scores.