Search papers, labs, and topics across Lattice.
Graduate School of Artificial Intelligence, POSTECH
2
0
4
Harness-aware post-training can drastically improve LLM agent performance, especially when facing shifting tool environments, revealing a key design dimension often overlooked in AI systems.
Current concept unlearning methods for diffusion models are easily bypassed by paraphrasing prompts, but PURE leverages cross-attention activations to achieve significantly better erasure.
