Proposes SCSuff metric for evaluating LLM explanation sufficiency via model-generated alternative inputs, showing explanations are typically insufficient and predictable from hidden states.
I ncorporating R esidual and N ormalization L ayers into A nalysis of M asked L anguage M odels
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
Positional encodings shift ViT representations toward index-anchored spatial structure, improving robustness to distribution shifts as quantified by a new SSDC metric.
citing papers explorer
-
What LLMs explain is not what they believe: Evaluating explanation sufficiency under models' own input beliefs
Proposes SCSuff metric for evaluating LLM explanation sufficiency via model-generated alternative inputs, showing explanations are typically insufficient and predictable from hidden states.
-
Positional Encodings Anchor Spatial Structure in Vision Transformers: A Geometric Perspective on Robustness
Positional encodings shift ViT representations toward index-anchored spatial structure, improving robustness to distribution shifts as quantified by a new SSDC metric.