We have 5 of 31 papers
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
INT8 is broadly safe for clinical deployment, while INT4 safety must be assessed per-model and per-task, and that safety alignment is determined primarily by instruction tuning rather than clinical domain adaptation.
The Cognitive Synergy Framework is introduced, a methodology for generating highquality humor data inspired by psychological theories of humor, and model variants significantly outperform larger instruction-tuned baselines and achieve top-tier open-weight performance while remaining competitive with frontier proprietar...
A reasoning-aware compression framework is presented that benchmarks quantization conditions across five reasoning benchmarks, GSM8K, FOLIO, MATH-500, ProofWriter, and MuSiQue, with hardware-level GPU energy measurement, and finds vulnerability is task-dependent and selective compression achieves Pareto-optimal points...
This work introduces the first information retrieval benchmark resource for Ehugbo, a high-quality parallel multimodal corpus constructed from a high-quality parallel multimodal corpus that reveals the ''Alignment Gap'', where African-centric foundation models that excel at linguistic familiarity with Igbo achieve <5%...
A large LLM-capability gap between the two languages is confirmed, and data augmentation experiments across three encoder models show that LLM-generated text consistently hurts downstream NER tasks while producing mixed effects on POS tagging, motivating careful language-specific IR evaluation.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
