LOCUS: Task-Aware Low-Rank Post-Training for Token-Efficient Language Generation

Excerpt

arXiv:2609.11739v1 Announce Type: new Abstract: Large language model serving costs scale directly with output sequence length, yet standard preference alignment often inflates response verbosity without improving utility. We study whether the parameterization of post-training updates affects generation length: low-rank subspaces alter sequence length without modifying the alignment loss. We present LOCUS, a method that selects a task-aware low-rank adaptation subspace to minimize output-token co