Showing 2026Show all
2 papers · 1 filter
cs.AI2026
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express
Lovedeep Gondara, Jekaterina Novikova, Sagarika Suresh +2
When an LLM judge filters instruction-tuning data, its scoring format (binary verdict, 1--5 rating, 0--100 score, or comparison) is rarely justified or ablated. Holding the candida…
cs.LG2026
A Coin Flip Per Token: Bernoulli Sparse Steering of Large Language Models
Nima Eshraghi, Lovedeep Gondara, Yuqing Huang +5
Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without task-specific fine-tuning, but standard methods apply the steering si…