7 papers
Spatial Supervision Without Attribution Optimization: Improving Post-Hoc Class Activation Maps via Box-Guided Evidence Routing
Wenhao Liang, Liangwei Nathan Zheng, Lin Yue +3
Post-hoc class activation maps (CAMs) are a standard tool for inspecting the evidence behind an image classifier's predictions, yet nothing in ordinary training encourages these ma…
Routing Probes Can Improve Without New Information: An Exact-Null Audit of Uncertainty Beyond Model Outputs
Wenhao Liang, Lin Yue, Wei Emma Zhang +3
Routing signals of modern vision transformers -- expert gates, attention-residual weights and halting scores -- often improve probes that predict whether the model is correct, and…
Signal-Routed Temperature Scaling: Low-Capacity Risk-Conditioned Calibration for Small Validation Budgets
Wenhao Liang, Liangwei Nathan Zheng, Lin Yue +4
When a classifier is recalibrated from only a few thousand held-out examples, the capacity of the calibration map becomes a statistical design choice rather than a purely architect…
How Much Can Reliability Drift Under a Fixed Confidence Distribution?
Wenhao Liang, Lin Yue, Wei Emma Zhang +3
A classifier's conditional accuracy can change while its confidence distribution stays exactly the same. We study the worst-case movement of the reliability relation under covariat…
Probing Routing-Conditional Calibration in Attention-Residual Transformers
Wenhao Liang, Lin Yue, Wei Emma Zhang +4
Post-hoc calibration is usually evaluated as a function of logits or softmax confidence alone, even as routing-augmented architectures increasingly accompany predictions with sampl…
Neural Tractability via Structure: Learning-Augmented Algorithms for Graph Combinatorial Optimization
Jialiang Li, Weitong Chen, Mingyu Guo
Neural solvers provide fast solutions to graph combinatorial optimization problems, but training alone does not guarantee solution quality. Exact methods guarantee optimality but c…
Calibration Attention: Learning Reliability-Aware Representations for Vision Transformers
Wenhao Liang, Wei Emma Zhang, Lin Yue +4
Most calibration methods operate at the logit level, implicitly assuming that miscalibration can be corrected without changing the underlying representation. We challenge this assu…