Search papers, labs, and topics across Lattice.
Affiliation:
7
0
11
Standard token-level SAEs squander their sparsity budgets memorizing punctuation and syntax; training on pooled multi-token chunks forces dictionaries to capture true, steerable semantic concepts and latent reasoning.
Decoupling task synthesis from on-policy rollouts solves the learning-signal saturation bottleneck in terminal agents, boosting long-horizon RL performance by up to 18 percentage points on Terminal-Bench 2.1.
A 440MB multilingual translation model now rivals commercial APIs, opening the door for performant on-device translation.
SkillSynth's skill graph approach lets you explicitly control the diversity of execution trajectories during terminal task synthesis, leading to more effective agent training.
On-policy data generation closes the training distribution gap and unlocks +2.54 performance gains at 128K context lengths, proving that LLMs learn best from data that evolves with their capabilities.
Forget rewriting single web pages – WRAP++ unlocks a 10x data scale-up and boosts LLM knowledge by synthesizing QA pairs that demand reasoning across multiple linked documents.
