Source author record

Dan Peng

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

ResearcherUnclaimed source record

Computer Vision Machine Learning physics.soc-ph Computation and Language Neural and Evolutionary Computing nlin.AO Social and Information Networks

Catalog footprint

What is connected

5works

7topics

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

SMMBench: A Benchmark for Source-Distributed Multimodal Agent Memory

Existing benchmarks for multimodal memory reasoning largely evaluate systems within pre-assembled contexts, but under-evaluate whether agents can use evidence distributed across independently originated sources. We argue that source-distributed memory composition is an important and under-examined bottleneck in multimodal agent memory, especially when relevant evidence is fragmented across heterogeneous artifacts such as conversations, profiles, screenshots, tables, images, and documents. To address this gap, we introduce Source-distributed Multimodal Memory Benchmark(SMMBench), which measures whether agents can retrieve, align, and compose multimodal evidence scattered across multiple sources rather than reason within a single curated context. SMMBench evaluates four core capabilities: (1) cross-source multimodal reasoning; (2) conflict resolution; (3) preference reasoning; (4) memory-grounded action prediction. The benchmark contains 1877 samples grounded in 264 sources. Experiments on representative memory-style and retrieval-based baselines show that current systems still struggle on these capabilities, positioning source-distributed multimodal memory as an important and still under-evaluated challenge for multimodal agents. Our data are available at https://huggingface.co/datasets/HuacanChai/SMMBench.

preprint2021arXiv

Organization of cooperation in fractal structures

It is known that the small-world structure constitutes sufficient conditions to sustain cooperation and thus enhances cooperation. On the contrary, the network with a very long average distance is usually thought of as suppressing the emergence of the cooperation. In this paper we show that the fractal structure, of which the average distance is very long, does not always play a negative role in the organization of cooperation. Compared to regular networks, the fractal structure might even facilitate the emergence of cooperation. This mainly depends on the existence of locally compact clusters. The sparse inter-connection between these clusters constructs an asymmetric barrier that the defection strategy is almost impossible to cross, but the cooperation strategy has a not too small chance. More generally, the network need not to be a standard fractal, as long as such structures exist. In turn, when this typical structure is absent, the fractal structure will also suppress the emergence of the cooperation, such as the fractal configuration obtained by diluting a random tree-like network. Our findings also clarify some contradictions in the previous studies, and suggest that both removing and inserting links from/into a regular network can enhance cooperation.

preprint2020arXiv

Decoder-free Robustness Disentanglement without (Additional) Supervision

Adversarial Training (AT) is proposed to alleviate the adversarial vulnerability of machine learning models by extracting only robust features from the input, which, however, inevitably leads to severe accuracy reduction as it discards the non-robust yet useful features. This motivates us to preserve both robust and non-robust features and separate them with disentangled representation learning. Our proposed Adversarial Asymmetric Training (AAT) algorithm can reliably disentangle robust and non-robust representations without additional supervision on robustness. Empirical results show our method does not only successfully preserve accuracy by combining two representations, but also achieve much better disentanglement than previous work.

preprint2020arXiv

Structure Matters: Towards Generating Transferable Adversarial Images

Recent works on adversarial examples for image classification focus on directly modifying pixels with minor perturbations. The small perturbation requirement is imposed to ensure the generated adversarial examples being natural and realistic to humans, which, however, puts a curb on the attack space thus limiting the attack ability and transferability especially for systems protected by a defense mechanism. In this paper, we propose the novel concepts of structure patterns and structure-aware perturbations that relax the small perturbation constraint while still keeping images natural. The key idea of our approach is to allow perceptible deviation in adversarial examples while keeping structure patterns that are central to a human classifier. Built upon these concepts, we propose a \emph{structure-preserving attack (SPA)} for generating natural adversarial examples with extremely high transferability. Empirical results on the MNIST and the CIFAR10 datasets show that SPA exhibits strong attack ability in both the white-box and black-box setting even defenses are applied. Moreover, with the integration of PGD or CW attack, its attack ability escalates sharply under the white-box setting, without losing the outstanding transferability inherited from SPA.

preprint2014arXiv

Punctuated equilibrium dynamics in human communications

A minimal model based on individual interactions is proposed to study the non-Poisson statistical properties of human behavior: individuals in the system interact with their neighbors, the probability of an individual acting correlates to its activity, and all individuals involved in action will change their activities randomly. The model creates rich non-Poisson spatial-temporal properties in the activities of individuals, in agreement with the patterns of human communication behaviors. Our findings provide insight into various human activities, embracing a range of realistic social interacting systems, particularly, intriguing bimodal phenomenons. This model bridges priority queues and punctuated equilibrium, and our modeling and analysis is likely to shed light on non-Poisson phenomena in many complex systems.