Tina Behnia

13 papers A* 2A 1B 2Journal 8
YearRankTypeTitle / Venue / Authors
2025 J jnl
Facts in Stats: Impacts of Pretraining Diversity on Language Model Generalization.
CoRR
Tina Behnia, Puneesh Deora, Christos Thrampoulidis
2025 J jnl
In-Context Occam's Razor: How Transformers Prefer Simpler Hypotheses on the Fly.
CoRR
Puneesh Deora, Bhavya Vasudeva, Tina Behnia, Christos Thrampoulidis
2024 J jnl
Implicit Geometry of Next-token Prediction: From Language Sparsity Patterns to Model Representations.
CoRR
Yize Zhao, Tina Behnia, Vala Vakilian, Christos Thrampoulidis
2024 B conf
Supervised Contrastive Representation Learning: Landscape Analysis with Unconstrained Features.
ISIT
Tina Behnia, Christos Thrampoulidis
2024 J jnl
Supervised Contrastive Representation Learning: Landscape Analysis with Unconstrained Features.
CoRR
Tina Behnia, Christos Thrampoulidis
2024 A* conf
Symmetric Neural-Collapse Representations with Supervised Contrastive Loss: The Impact of ReLU and Batching.
ICLR
Ganesh Ramachandra Kini, Vala Vakilian, Tina Behnia, Jaidev Gill, Christos Thrampoulidis
2023 A conf
On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data.
AISTATS
Tina Behnia, Ganesh Ramachandra Kini, Vala Vakilian, Christos Thrampoulidis
2023 J jnl
On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data.
CoRR
Tina Behnia, Ganesh Ramachandra Kini, Vala Vakilian, Christos Thrampoulidis
2023 J jnl
Supervised-Contrastive Loss Learns Orthogonal Frames and Batching Matters.
CoRR
Ganesh Ramachandra Kini, Vala Vakilian, Tina Behnia, Jaidev Gill, Christos Thrampoulidis
2022 A* conf
Imbalance Trouble: Revisiting Neural-Collapse Geometry.
NeurIPS
Christos Thrampoulidis, Ganesh Ramachandra Kini, Vala Vakilian, Tina Behnia
2022 J jnl
Imbalance Trouble: Revisiting Neural-Collapse Geometry.
CoRR
Christos Thrampoulidis, Ganesh R. Kini, Vala Vakilian, Tina Behnia
2022 B conf
On how to avoid exacerbating spurious correlations when models are overparameterized.
ISIT
Tina Behnia, Ke Wang, Christos Thrampoulidis
2022 J jnl
On how to avoid exacerbating spurious correlations when models are overparameterized.
CoRR
Tina Behnia, Ke Wang, Christos Thrampoulidis