optimization · 32 topics

nvidias sol pi slashes coding agent token use by 49

z image turbo

pruning llms like a physicist

googles dream rsi cuts ai discovery agent calls by 162x using replay

making knowledge distillation cheap enough

fable 5 vs gpt 5 6 sol np hard

model routing is simple until it isnt

why specialization is inevitable

2026 06 19 papers 2602 11557

2026 05 29 papers 2605.27375

parameter golf ai assisted research

2026 04 26 papers what actually breaks when you try to scale vehicle routing to 1m stops

optimizing the optimizer language models discover faster molecular relaxation

drift contract spectral updates

dstar straggler tolerant and byzantine resilient distributed sgd

ppo critic value flattening sp3o

optimal model activation policies for inference networks of large languages

saturation aware advantage reweighting

flowadam geometry aware soft momentum injection

guiding llms with gp evolved knowledge for project scheduling

kernelbrain coarse to fine budget aware search for agentic gpu kernel optimization

evothink evolving thinking in large reasoning models

muse representation geometry of muon beyond normalized momentum

polestar drift aware cache calibration

decoupling constraints from two directions

conformal predictive programming

dynamics fine tuning llms

when does learning pay off

msc cma es

lora ema generalization 2025

2605.19269

2605.19008