arXiv subjects

David Guo

Publications and source records attributed to David Guo.

6 recordsLinked to original sources

arXiv2026

Uniformly Almost Flatness and Solubility in Finitely Generated Groups

Tointon and the author conjectured that, for a finitely generated residually finite group, virtual nilpotence is equivalent to the condition that the diameters of its finite coset spaces admit a uniform polynomial lower bound in terms of their sizes. We first verify this conjecture for the class of finitely generated soluble groups. We then prove that this polynomial lower bound condition implies that the group has a finite-index subgroup whose finite quotients are all soluble. An immediate consequence of these two results is the verification of the conjecture for finitely generated linear groups. In addition, we establish the same conclusion for certain finitely generated abelian-by-cyclic groups under the weaker assumption that their finite quotients satisfy this polynomial lower bound condition.

math.GR↗

arXiv2026

Evaluating Scene-based In-Situ Item Labeling for Immersive Conversational Recommendation

The growing ubiquity of Extended Reality (XR) is driving Conversational Recommendation Systems (CRS) toward visually immersive experiences. We formalize this paradigm as Immersive CRS (ICRS), where recommended items are highlighted directly in the user's scene-based visual environment and augmented with in-situ labels. While item recommendation has been widely studied, the problem of how to select and evaluate which information to present as immersive labels remains an open problem. To this end, we introduce a principled categorization of information needs into explicit intent satisfaction and proactive information needs and use these to define novel evaluation metrics for item label selection. We benchmark IR-, LLM-, and VLM-based methods across three datasets and ICRS scenarios: fashion, movie recommendation, and retail shopping. Our evaluation reveals three important limitations of existing methods: (1) they fail to leverage scenario-specific information modalities (e.g., visual cues for fashion, meta-data for retail), (2) they present redundant information that is visually inferable, and (3) they poorly anticipate users' proactive information needs from explicit dialogue alone. In summary, this work provides both a novel evaluation paradigm for in-situ item labeling in ICRS and highlights key challenges for future work.

cs.IR↗

arXiv2025

VOGUE: A Multimodal Dataset for Conversational Recommendation in Fashion

Multimodal conversational recommendation has recently emerged as a promising paradigm for delivering personalized experiences through natural dialogue enriched by visual and contextual grounding. Yet currently available multimodal conversational recommendation datasets remain limited: existing resources either simulate conversations, omit user history or fail to collect sufficiently detailed feedback, which constrain the types of research and evaluation they support. To address these gaps we introduce VOGUE, a dataset of 60 human human dialogues containing 2100 granularly labeled utterances in realistic fashion shopping scenarios. Each dialogue is paired with a shared visual catalogue, item metadata, user fashion profiles and post conversation ratings from both users (Seekers) and recommenders (Assistants). This design enables rigorous evaluation of conversational inference, including not only alignment between predicted and ground truth preferences but also calibration against full rating distributions and comparison with explicit and implicit user satisfaction signals. Our analyses of VOGUE reveal distinctive dynamics of visually grounded dialogue, e.g. recommenders frequently recommend items simultaneously in feature based groups, which creates distinct conversational phases bridged by Seeker critiques and refinements. Benchmarking Multimodal Large Language Models against human Recommenders shows that while MLLMs approach human level alignment in aggregate they exhibit systematic distribution errors in reproducing human ratings and struggle to generalize preference inference beyond explicitly discussed items. These findings establish VOGUE as both a unique resource for studying multimodal conversational systems and a challenge dataset beyond the current recommendation capabilities of existing top tier multimodal foundation models such as GPT-5-mini and Gemini-2.5-Flash.

cs.IR↗

arXiv2024

Residually finite groups with uniformly almost flat quotients

We show that if all the finite coset spaces of a polycyclic group have diameter bounded uniformly below by a polynomial in their size then the group is virtually nilpotent. We obtain the same conclusion for a finitely generated residually torsion-free nilpotent group under the weaker assumption that the finite quotient groups have diameter bounded uniformly below by a polynomial in their size. This extends work of Khukhro and Valette.

math.GR↗

arXiv2024

The Conjugacy Ratio of Abelian-by-Cyclic Groups

Let $G = K \rtimes \langle t \rangle $ be a finitely generated group where $K$ is abelian and $\langle t\rangle$ is the infinite cyclic group. Let $ R $ be a finite symmetric subset of $K$ such that $S = \{ (r,1),(0,t^{\pm 1}) \mid r \in R \}$ is a generating set of $G$. We prove that the spherical conjugacy ratio, and hence the conjugacy ratio, of $G$ with respect to $S$ is $0$ unless $G$ is virtually abelian, confirming a conjecture of Ciobanu, Cox and Martino in this case. We also show that the Baumslag--Solitar group $\mathrm{BS}(1,2)$ has a one-sided F{\o}lner sequence $F$ such that the conjugacy ratio with respect to $F$ is non-zero, even though $\mathrm{BS}(1,2)$ is not virtually abelian. This is in contrast to two-sided F{\o}lner sequences, where Tointon showed that the conjugacy ratio with respect to a two-sided F{\o}lner sequence is positive if and only if the group is virtually abelian.

math.GR↗

arXiv2021

Preproduction Deploys: Cloud-Native Integration Testing

The microservice architecture for cloud-based systems is extended to not only require each loosely coupled component to be independently deployable, but also to provide independent routing for each component. This supports canary deployments, green/blue deployments and roll-back. Both ad hoc and system integration test traffic can be directed to components before they are released to production traffic. Front-end code is included in this architecture by using server-side rendering of JS bundles. Environments for integration testing are created with preproduction deploys side by side with production deploys using appropriate levels of isolation. After a successful integration test run, preproduction components are known to work with production precisely as it is. For isolation, test traffic uses staging databases that are copied daily from the production databases, omitting sensitive data. Safety and security concerns are dealt with in a targeted fashion, not monolithically. This architecture scales well with organization size; is more effective for integration testing; and is better aligned with agile business practices than traditional approaches.

cs.NI↗