Source author record

Po Hu

Po Hu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

math.AT Artificial Intelligence math.KT Computation and Language Computer Vision Machine Learning math.GT math.QA

Catalog footprint

What is connected

7works

8topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

Generalized Category Discovery under Domain Shifts: From Vision to Vision-Language Models

Generalized Category Discovery (GCD) aims to categorize unlabelled instances from both known and unknown classes by transferring knowledge from labelled data of known classes. Existing methods assume all data comes from a single domain, yet real-world unlabelled data often exhibits domain shifts alongside semantic shifts. We study GCD under domain shifts and propose three frameworks that adapt foundation models, ranging from self-supervised vision models to vision-language models. (i) HiLo disentangles domain and semantic features through multi-level feature extraction and mutual information minimization, combined with PatchMix augmentation and curriculum sampling. (ii) HLPrompt extends HiLo with semantic-aware spatial prompt tuning to suppress background and domain noise. (iii) VLPrompt leverages vision-language models via factorized textual prompts and cross-modal consistency regularization. The three methods share core design principles while operating on different foundation backbones, making them suitable for different deployment scenarios. Extensive experiments on synthetic corruptions and real-world multi-domain shifts demonstrate consistent improvements over strong baselines. Project page: https://visual-ai.github.io/hilo/

preprint2022arXiv

Triples-to-Text Generation with Reinforcement Learning Based Graph-augmented Neural Networks

Considering a collection of RDF triples, the RDF-to-text generation task aims to generate a text description. Most previous methods solve this task using a sequence-to-sequence model or using a graph-based model to encode RDF triples and to generate a text sequence. Nevertheless, these approaches fail to clearly model the local and global structural information between and within RDF triples. Moreover, the previous methods also face the non-negligible problem of low faithfulness of the generated text, which seriously affects the overall performance of these models. To solve these problems, we propose a model combining two new graph-augmented structural neural encoders to jointly learn both local and global structural information in the input RDF triples. To further improve text faithfulness, we innovatively introduce a reinforcement learning (RL) reward based on information extraction (IE). We first extract triples from the generated text using a pretrained IE model and regard the correct number of the extracted triples as the additional RL reward. Experimental results on two benchmark datasets demonstrate that our proposed model outperforms the state-of-the-art baselines, and the additional reinforcement learning reward does help to improve the faithfulness of the generated text.

preprint2015arXiv

D-structures and derived Koszul duality for unital operad algebras

Generalizing a concept of Lipshitz, Ozsváth and Thurs-ton from Bordered Floer homology, we define $D$-structures on algebras of unital operads, which can also be interpreted as a generalization of a seemingly unrelated concept of Getzler and Jones. This construction gives rise to an equivalence of derived categories, which can be thought of as a unital version of Koszul duality using non-unital Quillen homology. We also discuss a multi-sorted version of the construction, which provides a framework for unifying the known algebraic contexts of Koszul duality.

preprint2015arXiv

Field theories, stable homotopy theory and Khovanov homology

In this paper, we discuss two topics: first, we show how to convert 1+1-topological quantum field theories valued in symmetric bimonoidal categories into stable homotopical data, using a machinery by Elmendorf and Mandell. Then, we discuss, in this framework, two recent results (independent of each other) on refinements of Khovanov homology: our refinement into a module over the connective k-theory spectrum and a stronger result by Lipshitz and Sarkar refining Khovanov homology into a stable homotopy type.

preprint2014arXiv

Equivariant K-theory of compact Lie groups with involution

For a compact simply connected simple Lie group $G$ with an involution $α$, we compute the $G\rtimes \Z/2$-equivariant K-theory of $G$ where $G$ acts by conjugation and $\Z/2$ acts either by $α$ or by $g\mapsto α(g)^{-1}$. We also give a representation-theoretic interpretation of those groups, as well as of $K_G(G)$.

preprint2011arXiv

Topological Hermitian Cobordism

Extending our method for investigating Real cobordism (which was recently used by Hill, Hopkins and Ravenel in their solution of the Kervaire invariant 1 problem), we investigate the $RO(G)$-graded homotopy groups of a (non-complete) $\Z/2\times \Z/2$-equivariant spectrum called topological Hermitian cobordism. The methods of this paper may be useful in computing the homotopy groups of other $G$-equivariant spectra where $G\neq \Z/2$.

preprint2002arXiv

On Real-oriented Johnson-Wilson cohomology

Answering a question of W. S. Wilson, I introduce a Z/2-equivariant Atiyah-Real analogue of Johnson-Wilson cohomology theory BP<n>, whose coefficient ring is the =< n-chromatic part of Landweber's Real cobordism ring.