Showing cs.CLShow all
2 papers · 1 filter
cs.CL2026
Arabic Safety Alignment as Selective Refusal: An Empirical Study of SFT, DPO, and Guard Calibration
Mohamad Zbib, Ammar Mohanna
Arabic large language models must refuse harmful prompts without over-refusing benign or sensitive prompts, yet a single refusal rate hides this trade-off. We evaluate it using ben…
cs.CL2026
TAPS: Task Aware Proposal Distributions for Speculative Sampling
Mohamad Zbib, Mohamad Bazzi, Ammar Mohanna +2
Speculative decoding accelerates autoregressive generation by letting a lightweight draft model propose future tokens that a larger target model then verifies in parallel. In pract…