Author

Peng-Ji He

h-index 0 View on Semantic Scholar

We have 3 of 4 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

#artificial intelligence Preprint Aug 2026

HiPACE: Hierarchical Phase-Boundary Analysis and Controlled Evaluation of Feature Absorption in Sparse Autoencoders

Sparse autoencoders (SAEs) decompose LLM activations into sparse dictionary atoms, so that each distinct concept gets its own feature. One recurring behavior complicates this premise: feature absorption, in which a parent concept and its children--fruit and {apple, banana, pear}, say--collapse into a shared family dire...

Jin-Yuan Zhang, Peng-Ji He, Yin Yuan et al. · 0 citations

Preprint Aug 2026

Neither Precision Nor Architecture Alone: Controlled Tests of Failure Remedies for Physics-Informed Neural Networks

Physics-Informed Neural Networks (PINNs) frequently fail on stiff or advection-dominated PDEs, and two recent accounts offer competing remedies: switching from FP32 to FP64 to repair an L-BFGS stopping artifact, or replacing the MLP with a state-space-model (SSM) backbone plus sub-sequence alignment to counter architec...

Jin-Yuan Zhang, Peng-Ji He, He-Long Hu et al. · 0 citations

#artificial intelligence Preprint Aug 2026

Attention Sensitivity Is Not Enough: Dissociating Attention-Level and Behavioural In-Context Learning under Fine-Tuning

The main lesson is diagnostic: attention-level ICL proxies earn their place as training targets only after validation against behavioural gaps, and how far they can be trusted once it is optimised is asked.

Jin-Yuan Zhang, Peng-Ji He, He-Long Hu et al. · 0 citations