Search papers, labs, and topics across Lattice.
4
0
3
7
Labels can mislead researchers about task demands, as benchmarks with the same label can vary dramatically in their requirements and agent performance.
Targeted post-training interventions can boost code generation performance by over 6 points on key benchmarks, challenging the notion that larger models are always better.
LLMs struggle with code deletion, often opting for workarounds that compromise code maintainability, revealing a significant training gap in their editing capabilities.
Fine-tuning a small language model with MindForge's innovative training environments boosts its performance to rival much larger models in software engineering tasks.
