arXiv subjects

Eric Tang

Publications and source records attributed to Eric Tang.

13 recordsLinked to original sources

arXiv2026

Filling Positions Without Transfers: Screening on Outside Options

A designer offers vertically-differentiated positions to agents in the absence of transfers. Agents have private outside options and may reject their offers ex-post. The designer has preferences over the quantity of agents who accept each position. We show that under a general condition on the distribution of outside options, an optimal mechanism for the designer offers all agents an identical lottery, and we characterize this mechanism. When our condition does not hold, the optimal mechanism may require screening agents by offering a menu of distinct lotteries. Our results follow from a decomposition of agents' participation probabilities in any feasible mechanism.

econ.TH↗

arXiv2025

SkyRL-Agent: Efficient RL Training for Multi-turn LLM Agent

We introduce SkyRL-Agent, a framework for efficient, multi-turn, long-horizon agent training and evaluation. It provides efficient asynchronous dispatching, lightweight tool integration, and flexible backend interoperability, enabling seamless use with existing RL frameworks such as SkyRL-train, VeRL, and Tinker. Using SkyRL-Agent, we train SA-SWE-32B, a software engineering agent trained from Qwen3-32B (24.4% Pass@1) purely with reinforcement learning. We introduce two key components: an optimized asynchronous pipeline dispatcher that achieves a 1.55x speedup over naive asynchronous batching, and a tool-enhanced training recipe leveraging an AST-based search tool to facilitate code navigation, boost rollout Pass@K, and improve training efficiency. Together, these optimizations enable SA-SWE-32B to reach 39.4% Pass@1 on SWE-Bench Verified with more than 2x cost reduction compared to prior models reaching similar performance. Despite being trained solely on SWE tasks, SA-SWE-32B generalizes effectively to other agentic tasks, including Terminal-Bench, BrowseComp-Plus, and WebArena. We further demonstrate SkyRL-Agent's extensibility through case studies on deep research, computer use, and memory agents, each trained using a different training backend.

cs.AI↗

arXiv2025

Private From Whom? Minimal Information Leakage in Auctions

In many auctions, bidders may value keeping their private information hidden from the auctioneer or other bidders. Yet information must be conveyed to conduct an auction. Among deterministic bilateral communication protocols, revealing less information to bidders requires revealing more information to the auctioneer, and vice versa. A protocol implementing a given social choice rule is on the Privacy Frontier if no alternative protocol reveals less to both bidders and the auctioneer. For first-price auctions, the descending protocol and the sealed-bid protocol are on the Privacy Frontier. For second-price auctions, the ascending protocol and the ascending-join protocol are on the Privacy Frontier, but the sealed-bid protocol is not. We provide sufficient conditions for a protocol to be on the Privacy Frontier and devise alternative protocols allowing a designer to flexibly balance different dimensions of privacy.

econ.TH↗

arXiv2025

Empirical Computation: Prompting versus Programming

Large Language Models (LLM) can solve *any* computational problem *without* an algorithm in a runtime *independent* of the computational complexity of that problem. Instead of specifying precisely how to solve problem instance using *programming*, we ask an LLM to solve the problem instance using *prompting*. Outputs are sampled from a distribution rather than generated procedurally. In this vision paper, we explore the challenges and opportunities of this new form of computation and observe that its capabilities and limits *cannot* be understood within the classic, rationalist framework of computation. Hence, we appeal to the software engineering (SE) community to develop the foundations and techniques required to analyze the properties of this "empirical computation" as it generates solutions to computational problems: How can we analyze and improve the correctness of LLMs solving a computational problem in the general, in the problem-specific, or in the instance-specific? What are the properties and fundamental limits of empirical computation? This paper aims to establish empirical computation as a field in SE that is timely and rich with interesting problems.

cs.SE↗

arXiv2025

Cued to Queue: Information in Waiting-Line Auctions

We study the effect of providing information to agents who queue before a scarce good is distributed at a fixed time. Many information policies reveal "sudden bad news," when agents learn the queue is longer than previously believed. Sudden bad news causes assortative inefficiency by prompting multiple agents to simultaneously join the queue. If the value distribution has an increasing (decreasing) hazard rate, information policies that release sudden bad news increase (decrease) total surplus, relative to releasing no information. If agents incur entry costs and the hazard rate is decreasing, the optimal policy reveals only when the queue is full.

econ.TH↗

arXiv2025

LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!

Large reasoning models (LRMs) tackle complex reasoning problems by following long chain-of-thoughts (Long CoT) that incorporate reflection, backtracking, and self-validation. However, the training techniques and data requirements to elicit Long CoT remain poorly understood. In this work, we find that a Large Language model (LLM) can effectively learn Long CoT reasoning through data-efficient supervised fine-tuning (SFT) and parameter-efficient low-rank adaptation (LoRA). With just 17k long CoT training samples, the Qwen2.5-32B-Instruct model achieves significant improvements on a wide range of math and coding benchmarks, including 56.7% (+40.0%) on AIME 2024 and 57.0% (+8.1%) on LiveCodeBench, competitive to the proprietary o1-preview model's score of 44.6% and 59.1%. More importantly, we find that the structure of Long CoT is critical to the learning process, whereas the content of individual reasoning steps has minimal impact. Perturbations affecting content, such as training on incorrect samples or removing reasoning keywords, have little impact on performance. In contrast, structural modifications that disrupt logical consistency in the Long CoT, such as shuffling or deleting reasoning steps, significantly degrade accuracy. For example, a model trained on Long CoT samples with incorrect answers still achieves only 3.2% lower accuracy compared to training with fully correct samples. These insights deepen our understanding of how to elicit reasoning capabilities in LLMs and highlight key considerations for efficiently training the next generation of reasoning models. This is the academic paper of our previous released Sky-T1-32B-Preview model. Codes are available at https://github.com/NovaSky-AI/SkyThought.

cs.AI↗

arXiv2024

Streaming Detection of Queried Event Start

Robotics, autonomous driving, augmented reality, and many embodied computer vision applications must quickly react to user-defined events unfolding in real time. We address this setting by proposing a novel task for multimodal video understanding-Streaming Detection of Queried Event Start (SDQES). The goal of SDQES is to identify the beginning of a complex event as described by a natural language query, with high accuracy and low latency. We introduce a new benchmark based on the Ego4D dataset, as well as new task-specific metrics to study streaming multimodal detection of diverse events in an egocentric video setting. Inspired by parameter-efficient fine-tuning methods in NLP and for video tasks, we propose adapter-based baselines that enable image-to-video transfer learning, allowing for efficient online video modeling. We evaluate three vision-language backbones and three adapter architectures on both short-clip and untrimmed video settings.

cs.CV↗

arXiv2024

Understanding LLM Embeddings for Regression

With the rise of large language models (LLMs) for flexibly processing information as strings, a natural application is regression, specifically by preprocessing string representations into LLM embeddings as downstream features for metric prediction. In this paper, we provide one of the first comprehensive investigations into embedding-based regression and demonstrate that LLM embeddings as features can be better for high-dimensional regression tasks than using traditional feature engineering. This regression performance can be explained in part due to LLM embeddings over numeric data inherently preserving Lipschitz continuity over the feature space. Furthermore, we quantify the contribution of different model effects, most notably model size and language understanding, which we find surprisingly do not always improve regression performance.

cs.LG↗

arXiv2022

How Would The Viewer Feel? Estimating Wellbeing From Video Scenarios

In recent years, deep neural networks have demonstrated increasingly strong abilities to recognize objects and activities in videos. However, as video understanding becomes widely used in real-world applications, a key consideration is developing human-centric systems that understand not only the content of the video but also how it would affect the wellbeing and emotional state of viewers. To facilitate research in this setting, we introduce two large-scale datasets with over 60,000 videos manually annotated for emotional response and subjective wellbeing. The Video Cognitive Empathy (VCE) dataset contains annotations for distributions of fine-grained emotional responses, allowing models to gain a detailed understanding of affective states. The Video to Valence (V2V) dataset contains annotations of relative pleasantness between videos, which enables predicting a continuous spectrum of wellbeing. In experiments, we show how video models that are primarily trained to recognize actions and find contours of objects can be repurposed to understand human preferences and the emotional content of videos. Although there is room for improvement, predicting wellbeing and emotional response is on the horizon for state-of-the-art models. We hope our datasets can help foster further advances at the intersection of commonsense video understanding and human preference learning.

cs.CV↗

arXiv2022

On the Existence of Balanced Generalized de Bruijn Sequences

A balanced generalized de Bruijn sequence with parameters $(n,l,k)$ is a cyclic sequence of $n$ bits such that (a) the number of 0's equals the number of 1's, and (b) each substring of length $l$ occurs at most $k$ times. We determine necessary and sufficient conditions on $n,l$, and $k$ for the existence of such a sequence.

math.CO↗

arXiv2021

Imaging reconfigurable molecular concentration on a graphene field-effect transistor

The spatial arrangement of adsorbates deposited onto a clean surface in vacuum typically cannot be reversibly tuned. Here we use scanning tunneling microscopy to demonstrate that molecules deposited onto graphene field-effect transistors exhibit reversible, electrically-tunable surface concentration. Continuous gate-tunable control over the surface concentration of charged F4TCNQ molecules was achieved on a graphene FET at T = 4.5K. This capability enables precisely controlled impurity doping of graphene devices and also provides a new method for determining molecular energy level alignment based on the gate-dependence of molecular concentration. The gate-tunable molecular concentration can be explained by a dynamical molecular rearrangement process that reduces total electronic energy by maintaining Fermi level pinning in the device substrate. Molecular surface concentration in this case is fully determined by the device back-gate voltage, its geometric capacitance, and the energy difference between the graphene Dirac point and the molecular LUMO level.

cond-mat.mes-hall↗

arXiv2021

Validation of image systems simulation technology using a Cornell Box

We describe and experimentally validate an end-to-end simulation of a digital camera. The simulation models the spectral radiance of 3D-scenes, formation of the spectral irradiance by multi-element optics, and conversion of the irradiance to digital values by the image sensor. We quantify the accuracy of the simulation by comparing real and simulated images of a precisely constructed, three-dimensional high dynamic range test scene. Validated end-to-end software simulation of a digital camera can accelerate innovation by reducing many of the time-consuming and expensive steps in designing, building and evaluating image systems.

eess.IV↗

arXiv2021

Measuring Mathematical Problem Solving With the MATH Dataset

Many intellectual endeavors require mathematical problem solving, but this skill remains beyond the capabilities of computers. To measure this ability in machine learning models, we introduce MATH, a new dataset of 12,500 challenging competition mathematics problems. Each problem in MATH has a full step-by-step solution which can be used to teach models to generate answer derivations and explanations. To facilitate future research and increase accuracy on MATH, we also contribute a large auxiliary pretraining dataset which helps teach models the fundamentals of mathematics. Even though we are able to increase accuracy on MATH, our results show that accuracy remains relatively low, even with enormous Transformer models. Moreover, we find that simply increasing budgets and model parameter counts will be impractical for achieving strong mathematical reasoning if scaling trends continue. While scaling Transformers is automatically solving most other text-based tasks, scaling is not currently solving MATH. To have more traction on mathematical problem solving we will likely need new algorithmic advancements from the broader research community.

cs.LG↗