Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillation

Documenti analoghi

• CARE: Decoding Time Safety Alignment via Rollback and Introspection Intervention
  di: Hu, Xiaomeng, et al.
  Pubblicazione: (2025)
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
  di: Qiu, Zihan, et al.
  Pubblicazione: (2025)
• Finding Challenging Metaphors that Confuse Pretrained Language Models
  di: Li, Yucheng, et al.
  Pubblicazione: (2024)
• Predicting Rewards Alongside Tokens: Non-disruptive Parameter Insertion for Efficient Inference Intervention in Large Language Model
  di: Yuan, Chenhan, et al.
  Pubblicazione: (2024)
• Vision Language Models are Confused Tourists
  di: Irawan, Patrick Amadeus, et al.
  Pubblicazione: (2025)