arXiv subjects

Ben Snyder

Publications and source records attributed to Ben Snyder.

4 recordsLinked to original sources

arXiv2026

Clustering of z~6.6 Quasars and [O III] Emitters Constrains Host Halo Masses and Duty Cycles in 25 ASPIRE Fields

We use data from the JWST ASPIRE Wide Field Slitless Spectroscopy (WFSS) program to measure the auto-correlation function of [O,{\sc iii}] emitters at 5.3$<z<$7.0 and the quasar--[O,{\sc iii}] emitter cross-correlation around 25 ASPIRE quasars (6.51$<z<$6.82; $\langle z\rangle=6.6$). We use synthetic source injection to calibrate the selection function, which we combine with the large-volume FLAMINGO-10k simulation (2.8,cGpc box) to construct realistic mock observations. Our simulation-based approach captures nonlinear structure growth and scale-dependent bias on small scales and derives covariance matrices that include cosmic variance. The clustering yields correlation lengths of $r_0^{\rm GG}=4.7^{+0.4}{-0.5},h^{-1}$,cMpc for the [O,{\sc iii}] auto-correlation with fixed slope $\gamma{\rm GG}=1.8$, and $r_0^{\rm QG}=8.7^{+0.8}{-0.9},h^{-1}$,cMpc for the quasar--[O,{\sc iii}] cross-correlation with $\gamma{\rm QG}=2.0$. We infer $\log(M_{h,{\rm min}}^{[{\rm O,III}]}/M_\odot)=10.5^{+0.1}{-0.1}$ for [O,{\sc iii}] emitters and $\log(M{h,{\rm min}}^{\rm QSO}/M_\odot)=12.1^{+0.3}{-0.4}$ for quasars. These imply duty cycles of $2.5^{+1.0}{-0.8}$,per,cent for [O,{\sc iii}] emitters and $0.3^{+4.0}{-0.3}$,per,cent for quasars, corresponding to UV-bright lifetimes of $t{\rm Q}=2.6^{+30}_{-2.5}$,Myr (less than 10,per,cent of a Salpeter $e$-folding time). The results indicate that the observed UV-luminous phase contributes little to total SMBH growth, placing tight constraints on early black-hole formation.

astro-ph.GA↗

arXiv2024

Uncertainty-Guided Enhancement on Driving Perception System via Foundation Models

Multimodal foundation models offer promising advancements for enhancing driving perception systems, but their high computational and financial costs pose challenges. We develop a method that leverages foundation models to refine predictions from existing driving perception models -- such as enhancing object classification accuracy -- while minimizing the frequency of using these resource-intensive models. The method quantitatively characterizes uncertainties in the perception model's predictions and engages the foundation model only when these uncertainties exceed a pre-specified threshold. Specifically, it characterizes uncertainty by calibrating the perception model's confidence scores into theoretical lower bounds on the probability of correct predictions using conformal prediction. Then, it sends images to the foundation model and queries for refining the predictions only if the theoretical bound of the perception model's outcome is below the threshold. Additionally, we propose a temporal inference mechanism that enhances prediction accuracy by integrating historical predictions, leading to tighter theoretical bounds. The method demonstrates a 10 to 15 percent improvement in prediction accuracy and reduces the number of queries to the foundation model by 50 percent, based on quantitative evaluations from driving datasets.

cs.CV↗

arXiv2023

On Early Detection of Hallucinations in Factual Question Answering

While large language models (LLMs) have taken great strides towards helping humans with a plethora of tasks, hallucinations remain a major impediment towards gaining user trust. The fluency and coherence of model generations even when hallucinating makes detection a difficult task. In this work, we explore if the artifacts associated with the model generations can provide hints that the generation will contain hallucinations. Specifically, we probe LLMs at 1) the inputs via Integrated Gradients based token attribution, 2) the outputs via the Softmax probabilities, and 3) the internal state via self-attention and fully-connected layer activations for signs of hallucinations on open-ended question answering tasks. Our results show that the distributions of these artifacts tend to differ between hallucinated and non-hallucinated generations. Building on this insight, we train binary classifiers that use these artifacts as input features to classify model generations into hallucinations and non-hallucinations. These hallucination classifiers achieve up to $0.80$ AUROC. We also show that tokens preceding a hallucination can already predict the subsequent hallucination even before it occurs.

cs.CL↗

arXiv2021

Task-Oriented Dialogue as Dataflow Synthesis

We describe an approach to task-oriented dialogue in which dialogue state is represented as a dataflow graph. A dialogue agent maps each user utterance to a program that extends this graph. Programs include metacomputation operators for reference and revision that reuse dataflow fragments from previous turns. Our graph-based state enables the expression and manipulation of complex user intents, and explicit metacomputation makes these intents easier for learned models to predict. We introduce a new dataset, SMCalFlow, featuring complex dialogues about events, weather, places, and people. Experiments show that dataflow graphs and metacomputation substantially improve representability and predictability in these natural dialogues. Additional experiments on the MultiWOZ dataset show that our dataflow representation enables an otherwise off-the-shelf sequence-to-sequence model to match the best existing task-specific state tracking model. The SMCalFlow dataset and code for replicating experiments are available at https://www.microsoft.com/en-us/research/project/dataflow-based-dialogue-semantic-machines.

cs.CL↗