arXiv subjects

James Lyon

Publications and source records attributed to James Lyon.

8 recordsLinked to original sources

arXiv2025

UISim: An Interactive Image-Based UI Simulator for Dynamic Mobile Environments

Developing and testing user interfaces (UIs) and training AI agents to interact with them are challenging due to the dynamic and diverse nature of real-world mobile environments. Existing methods often rely on cumbersome physical devices or limited static analysis of screenshots, which hinders scalable testing and the development of intelligent UI agents. We introduce UISim, a novel image-based UI simulator that offers a dynamic and interactive platform for exploring mobile phone environments purely from screen images. Our system employs a two-stage method: given an initial phone screen image and a user action, it first predicts the abstract layout of the next UI state, then synthesizes a new, visually consistent image based on this predicted layout. This approach enables the realistic simulation of UI transitions. UISim provides immediate practical benefits for UI testing, rapid prototyping, and synthetic data generation. Furthermore, its interactive capabilities pave the way for advanced applications, such as UI navigation task planning for AI agents. Our experimental results show that UISim outperforms end-to-end UI generation baselines in generating realistic and coherent subsequent UI states, highlighting its fidelity and potential to streamline UI development and enhance AI agent training.

cs.CV↗

arXiv2024

Sample what you cant compress

For learned image representations, basic autoencoders often produce blurry results. Reconstruction quality can be improved by incorporating additional penalties such as adversarial (GAN) and perceptual losses. Arguably, these approaches lack a principled interpretation. Concurrently, in generative settings diffusion has demonstrated a remarkable ability to create crisp, high quality results and has solid theoretical underpinnings (from variational inference to direct study as the Fisher Divergence). Our work combines autoencoder representation learning with diffusion and is, to our knowledge, the first to demonstrate jointly learning a continuous encoder and decoder under a diffusion-based loss and showing that it can lead to higher compression and better generation. We demonstrate that this approach yields better reconstruction quality as compared to GAN-based autoencoders while being easier to tune. We also show that the resulting representation is easier to model with a latent diffusion model as compared to the representation obtained from a state-of-the-art GAN-based loss. Since our decoder is stochastic, it can generate details not encoded in the otherwise deterministic latent representation; we therefore name our approach "Sample what you can't compress", or SWYCC for short.

cs.LG↗

arXiv2017

Now Playing: Continuous low-power music recognition

Existing music recognition applications require a connection to a server that performs the actual recognition. In this paper we present a low-power music recognizer that runs entirely on a mobile device and automatically recognizes music without user interaction. To reduce battery consumption, a small music detector runs continuously on the mobile device's DSP chip and wakes up the main application processor only when it is confident that music is present. Once woken, the recognizer on the application processor is provided with a few seconds of audio which is fingerprinted and compared to the stored fingerprints in the on-device fingerprint database of tens of thousands of songs. Our presented system, Now Playing, has a daily battery usage of less than 1% on average, respects user privacy by running entirely on-device and can passively recognize a wide range of music.

cs.SD↗

arXiv2015

Isospin asymmetries in $B\to (K^*,ρ) γ/ l^+ l^-$ and $B \to K l^+ l^-$ in and beyond the Standard Model

We compute the isospin asymmetries in $B \to (K^*,ρ) γ$ and $B\to (K,K^*,ρ) l^+l^-$ for low lepton pair invariant mass $q^2$, within the Standard Model (SM) and beyond the SM (BSM) in a generic dimension six operator basis. Within the SM the CP-averaged isospin asymmetries for $B \to (K,K^*,ρ) ll$, between $1{\rm GeV}^2 \leq q^2 \leq 4m_c^2$, are predicted to be small (below 1.5%) though with significant cancellation. In the SM the non-CP averaged asymmetries for $B \to ρll$ deviate by $\approx \pm 5%$ from the CP-averaged ones. We provide physical arguments, based on resonances, of why isospin asymmetries have to decrease for large $q^2$ (towards the endpoint). Two types of isospin violating effects are computed: ultraviolet (UV) isospin violation due to differences between operators coupling to up and down quarks, and infrared (IR) isospin violation where a photon is emitted from the spectator quark and is hence proportional to the difference between the up- and down-quark charges. These isospin violating processes may be subdivided into weak annihilation (WA), quark loop spectator scattering (QLSS) and a chromomagnetic contribution. Furthermore we discuss generic selection rules based on parity and angular momentum for the $B \to Kll$ transition as well as specific selection rules valid for WA at leading order in the strong coupling constant. We clarify that the relation between the $K$ and the longitudinal part of the $K^*$ only holds for leading twist and for left-handed currents. In general the $B \to ρll$ and $B \to K^*ll$ isospin asymmetries are structurally different yet the closeness of $α_{\rm CKM}$ to ninety degrees allows us to construct a (quasi) null test for the SM out of the respective isospin symmetries. We provide an update on ${\cal B}(B^0 \to K^{*0}γ)/{\cal B}(B_s \to ϕγ)$ which is sensitive to WA.

hep-ph↗

arXiv2014

Resonances gone topsy turvy - the charm of QCD or new physics in $b \to s \ell^+ \ell^-$?

We investigate the interference pattern of the charm-resonances $Ψ(3370,4040,4160,4415)$ with the electroweak penguin operator $O_9$ in the branching fraction of $B^+\to K^+μμ$. For this purpose we extract the charm vacuum polarisation via a standard dispersion relation from BESII-data on $e^+e^-\to hadrons$. In the factorisation approximation (FA) the vacuum polarisation describes the interference fully non-perturbatively. The observed interference pattern by the LHCb collaboration is opposite in sign and and significantly enhanced as compared to the FA. A change of the FA-result by a factor of -2.5, which correspond to a 350%-corrections, results in a reasonable agreement with the data. This raises the question on the size of non-factorisable corrections which are colour enhanced but loop-suppressed. In the parton picture it is found that the corrections are of relative size ~-0.5 when averaged over the open charm-region which is far below -3.5 needed to explain the observed effect. We present combined fits to the BESII- and the LHCb-data, testing for effects beyond the Standard Model (SM)-FA. We cannot find any significant evidence of the parton estimate being too small due to cancellations between the individual resonances. It seems difficult to accommodate the LHCb-result in the standard treatment of the SM or QCD respectively. In the SM the effect can be described in a $q^2$-dependent shift of the Wilson coefficient combination $C^{eff}_9 + C^{' eff}_9$. We devise strategies to investigate the microscopic structure in future measurements. We show that the charm-resonance effects can accommodate the $B \to K^* ll$-anomalies (e.g. $P_5'$). Hence our findings indicate that the interpretation of the anomaly through a $Z'$-boson, mediating between $bs$ and $ll$ fields, is disfavoured. More generally our results motivate investigations into $b \to s\bar cc$-physics.

hep-ph↗

arXiv2013

Exclusive Chromomagnetism in heavy-to-light FCNCs

We compute matrix elements of the chromomagnetic operator, often denoted by $ {\cal O}_8$, between $B/D$-states and light mesons plus an off-shell photon by employing the method of light-cone sum rules (LCSR) at leading twist-2. These matrix elements are relevant for processes such as $B \to K^* l^+l^-$ and they can be seen as the analogues of the well-known penguin form factors $T_{1,2,3}$ and $f_T$. We find a large CP-even phase for which we give a long-distance (LD) interpretation. We compare our results to QCD factorisation for which the spectator photon emission is end-point divergent. The analytic structure of the correlation function used in our method admits a complex anomalous threshold on the physical sheet. The meaning and handling within the sum rule approach of the anomalous threshold is discussed.

hep-ph↗

arXiv2013

Heavy-to-light chromomagentic matrix element

We report the computation of the matrix element of the chromomagnetic operator of the flavour changing neutral current (FCNC)-type between a $B$- or $D$-meson state and a light hadron and off-shell photon. The computation is carried out by using the method of light-cone sum rules (LCSR). It is found that the matrix element exhibits a large strong phase for which we give a long distance interpretation. The analytic structure of the correlation function in use admits a complex anomalous threshold on the physical sheet, the meaning and handling of which within the sum rule approach is discussed. We compare our results to QCD factorisation for which spectator photon emission is end-point divergent.

hep-ph↗

arXiv2012

$A_{\textrm{CP}}[D_{(s)}^{0,+} \to V \gamma]$ from Large ${\cal O}_8$

CP-violation in $\Delta A_{\textrm {CP}} = -0.154(29)$, in the $D^0 \to \pi^+\pi^-/K^+K^-$ system, is established and its central value is one order of magnitude above the naive Standard Model (SM) estimate. It remains unclear whether this is due to currently incalculable strong interaction matrix elements or genuine new physics (NP) such as a shift in ${\cal O}_8$ with a weak phase. We show that the interference of the long-distance (LD) terms with the ${\cal O}_8$ matrix element can give rise to $A_{\text{CP}}^{D \to V \gamma} = \textrm{few} \cdot 10^{-3}$ (for reference values ${\textrm{Im}}[C_8^{NP}] \approx 10^{-3}$). In addition, it is pointed out that the ratio of left- to right-handed (photon polarisation) LD amplitudes is measurable in time-dependent CP (TDCP) asymmetries. We argue that both theory and experimental consideration favour weak annihilation (WA) as the dominant LD contribution. More definite progress could be achieved by either computing the radiative corrections to WA or the measurement of the charged modes $D^+_{(d,s)} \to (\rho,K^{*})^+ \gamma$ and $D_s \to \rho^+ \gamma$.

hep-ph↗