Search papers, labs, and topics across Lattice.
3
0
6
0
High Pass@1 masks a critical failure mode in frontier code models: models routinely ship bloated, unreviewable diffs for trivial bugs, an issue that scaling cannot solve but RL post-training can directly optimize.
Metacognitive regulation emerges as a key differentiator for deeper collaboration in human-AI interactions, reshaping our understanding of dialogue dynamics.
Optimal mixing of query translations can significantly boost retrieval performance, but English dominance complicates the landscape of multilingual querying.
