Search papers, labs, and topics across Lattice.
10
0
12
Static memory models are holding back performance鈥擯roteus reveals that incrementally activating memory can drastically enhance context retention and reduce interference.
LLMs can now tackle complex TCS proof generation with a benchmark that achieves over 90% accuracy in verification against human experts.
SSTQ cuts communication costs in federated learning while ensuring privacy, achieving optimal mean squared error scaling with minimal bit usage.
No randomized polynomial-time algorithm can overcome the condition-number barrier in sparse least-squares optimization, confirming a long-standing conjecture.
Nearly optimal dimension lower bounds reveal that single-vector embeddings for MAX-IP require significantly more dimensions than previously thought, closing a critical gap in the literature.
A new AI tool can catch 34% more mathematical errors in scientific papers, transforming the peer review landscape.
Language models can achieve remarkable improvements in learning and memory retention by mimicking the human processes of sleep and dreaming.
Recurrent models can now achieve Transformer-competitive performance on recall-intensive tasks, thanks to a simple memory caching mechanism that grows memory capacity with sequence length.
Gemini 3 Deep Think can now autonomously solve a majority of problems in a challenging math competition, signaling a leap in AI's mathematical reasoning capabilities.
Surprisingly, using only a single inner loop update in data mixing can lead to failure, and the optimal number of inner loop steps scales logarithmically with the parameter update budget.
