Cambridge NLIP Multimodal BabyLM 2025: decoder with token-wise dynamic gating, feature modulation, and channel attention
AI & ML interests
Multimodal BabyLM
Recent Activity
View all activity
Papers
View all Papers
Organization Card
LOOKING TO LEARN: Token-wise Dynamic Gating for Low-Resource Vision-Language Modelling
models
15
LookingtoLearn/channel_attention_model
Updated
LookingtoLearn/base_model
Updated
LookingtoLearn/ntp_lcg_objectives_model
Updated
LookingtoLearn/ntp_clip_objectives_model
Updated
LookingtoLearn/no_encoder_model
Updated
LookingtoLearn/mlp_encoder_model
Updated
LookingtoLearn/gate_soft_per_token
Updated
LookingtoLearn/gate_none_model
Updated
LookingtoLearn/gate_hard_per_token_model
Updated
LookingtoLearn/gate_hard_per_feature_model
Updated