Searcharxiv⌕ Search

arXiv subjects

Yichuan Wang

Publications and source records attributed to Yichuan Wang.

At least 19 recordsLinked to original sources

StreamEmbed: Real-Time Indexing from Pixels for Screenshot Retrieval

Screen history can serve as searchable visual memory for people and agents, but unlike an offline corpus, it must be encoded and indexed as each screenshot arrives. OCR-based indexing is fast but loses visual information; visual encoding takes over two seconds. Yet screens change incrementally: most pixels repeat within and across screenshots. This raises a question: do visual encoders need to recompute their representations from scratch for every new screenshot? We present StreamEmbed, a system for incremental visual embedding that exploits this redundancy. Guided solely by pixel differences, it combines spatial token pruning, which skips blocks that repeat a neighbor, with temporal KV reuse, which reuses cached keys and values for blocks unchanged across screenshots. A small learned adapter compensates for pruning and reuse to preserve fidelity to full-screenshot embeddings. On ScreenMemEval, our benchmark of 10,426 questions over 184 hours of screen histories from two humans and a computer-use agent, StreamEmbed sets a new accuracy-latency Pareto frontier on an M2 Pro GPU: it encodes up to 3.7x faster than full-screenshot encoding without sacrificing accuracy and improves accuracy over OCR-based indexing by up to 6.3 Recall@10 points at matched latency. A memory-efficient vector index stores a simulated year of screenshots in 0.13 GB of RAM. Together, these results make real-time indexing of screen history in its native visual form practical, laying a foundation for fully local visual assistants.

cs.CV↗

Low Probability of Target Detection in Wireless Sensing by Harnessing Reconfigurable-antenna based Reflectors

In this paper, we introduce low probability of target detection (LPTD) as a new security objective for wireless sensing systems, where the existence of a target is concealed from passive wardens. Unlike conventional communication security and covert communication, LPTD aims to protect the sensed target itself from malicious detection, rather than concealing transmitted signals or communication activities. Specifically, we propose a reconfigurable-antenna-based reflector-assisted target covertness scheme, where passive reflectors deployed around the target are harnessed as artificial scattering sources to confuse the warden. By adjusting their reflection coefficients, these reflectors reshape the local scattering environment and mask the target-induced echo. We then analyze the warden's optimal detection strategy and derive the false-alarm and missed-detection probabilities, followed by a closed-form expression for its minimum total detection error probability. Numerical results validate the analysis and show that sufficient cooperation among reflectors can increase the warden's minimum total detection error probability to approximately 0.94--0.96, making target detection close to random guessing.

cs.CR↗

GeoGuesser Bench: Benchmarking Vision-Language Agents on Interactive Geolocation

Vision-language model (VLM) agents that act in the open world need two abilities: world knowledge to interpret what they see, and exploration to find what they cannot yet see. We study how agents combine the two through interactive visual geolocation. GeoGuesser Bench contains 282 tasks in an offline Street View environment where agents rotate, zoom, and move between real panoramas. Its three task types shift the balance from knowledge to exploration: agents (1) report the coordinates of their starting location, (2) name the enclosing street, city, state, or country, or (3) find, read, and count nearby objects. The coordinate tasks in (1) also include 42 real GeoGuessr World Championship rounds, whose expert guesses provide a human reference. We evaluate 23 VLMs: the best accuracy is 62.5% on place naming and 35.0% on object finding, and the best agent averages 4352.4 GeoGuessr points on the championship rounds, against 4770.3 for the better expert in each round. Beyond these headline numbers, world knowledge and exploration are only weakly correlated across models: agents that localize well do not necessarily explore well. To understand where agents fail, we categorize their errors into vision, navigation, and tool-use failures. Further ablations show that the strongest agents, such as Claude Opus 4.8, still perform well with limited views, while weaker agents degrade; customizing the environment thus adjusts task difficulty. The environment's verifiable rewards and restorable state also support training, and our Planner-Worker-Compressor harness cuts planner tokens by 57-61% by compacting the interaction history. We hope GeoGuesser Bench and its environment serve as a high-quality foundation for simulating real-world exploration at scale.

cs.AI↗

DR Pixel: Page-Native Deep Research with Vision-Language Models

Complex documents such as scientific papers contain rich information in tables, figures, and equations; however, existing deep research (DR) agents typically reduce papers to OCR-extracted text. We instead ask whether vision-language models (VLMs) can enable page-native deep research: directly retrieving and reasoning over rendered document page images. We introduce DR Pixel, the first open-source end-to-end multimodal DR agent that uses rendered page images as its sole retrieval and reasoning unit, operating over a corpus of 1.06M pages. To evaluate this setting, we construct DRPixelBench, a benchmark on scientific papers requiring reasoning over text, figures, tables, and equations. Surprisingly, page-native DR already works well without training, outperforming OCR-based baselines across state-of-the-art models like GPT-5.4, Gemini-3.5-Flash, and Qwen3.6-27B, while requiring fewer tool calls. We next turn to DR training. A key challenge is that teacher trajectories for agentic supervised finetuning are noisy, mixing critical reasoning and action steps with redundant or failed attempts. To address this, we introduce counterfactual credit reweighting, a fine-grained credit assignment method that weights individual reasoning and action steps by their counterfactual contribution to the final answer. Our training improves retrieval recall by about 15 points and final accuracy by over 10 points on DRPixelBench, and likewise outperforms the untrained baseline on four existing benchmarks. On out-of-distribution ViDoRe-V3, DR Pixel comes within 1.8 points of an oracle that is given the gold pages directly. Our results suggest that the conventional OCR-then-retrieve pipeline is not only unnecessary but suboptimal: directly retrieving and reasoning over visual document pages makes complex DR both more effective and more efficient.

cs.CV↗

FrontierCS 2: A Benchmark for Computer Science R&D

Can LLM agents make contributions to frontier computer science R&D? We introduce FrontierCS 2, a benchmark of 48 open-ended tasks. We take a paper and the code its authors released and delete the paper's contribution. The deleted code becomes the human reference that sets the bar. Earlier open-ended benchmarks give each agent a problem statement and grade one program after the run. FrontierCS 2 places the agent inside a research environment. The agent inherits a working repository and may submit a patch at any point. For each patch, a judge returns a development-suite score but keeps the evaluator, the held-out final suite and the reference private, and grades the final submission on that suite against the reference. The tasks span 13 domains, from machine-learning systems and high-performance computing to databases, security and robotics. Beyond these 48 tasks, FrontierCS 2 keeps a private set of more than 100 tasks, to be announced in the future. We evaluate 12 agents. The best one is Claude Opus 5.5, passing 58% of its runs. No run passes on 16 tasks. A higher score is not the same as a pass: most runs that beat the reference on the combined final score still fall short on some metric or workload. On a shared capability scale for agents and tasks, the median task is roughly as difficult as what the strongest agent can handle.

cs.LG↗

EasyPPO: Stabilizing the Critic Is Key

A key strength of Proximal Policy Optimization (PPO) is its learned critic, which uses historical trajectories collected during reinforcement learning to estimate expected returns and reduce policy-gradient variance. However, we find that the critic is also a major source of instability in reinforcement learning for large language models (LLMs). We identify two critic failure modes that destabilize PPO. First, filtering truncated rollouts from both actor and critic shifts the policy objective to reward conditioned on completion, allowing truncation to increase even as conditional reward improves. Second, heterogeneous return noise can cause high-variance prompts to dominate critic updates in finite batches. We introduce EasyPPO to address these failures. Actor-only overlong filtering trains the critic on returns from both completed and truncated rollouts. Noise-normalized critic regression weights each prompt's critic loss by the inverse standard deviation of its sampled returns, balancing noise contributions across prompts. Moderately smaller critic mini-batches confine outlier influence to fewer rollouts during gradient clipping. Across continuous-reward coding on FrontierCS, binary-reward mathematical reasoning on AIME24, and multi-turn search on Search-R1, EasyPPO remains stable throughout the full training horizon and consistently outperforms vanilla PPO, VAPO, and HL-Gauss PPO. Its best validation scores show relative gains of 14.89%, 2.28%, and 9.47% over PPO, respectively.

cs.LG↗

Token Latency Fairness: Performance Isolation for Multi-Tenant LLM Serving

LLM serving is typically offered as a shared, multi-tenant service, where high-demand workloads from one client can cause latency SLO violations for others. Existing solutions for performance isolation equalize client throughput in the long run, for example through queueing and batching fairness. However, these approaches do not provide latency isolation guarantees; as a result, well-behaved clients can still experience significant degradation to their token-level latencies. In this paper, we present FairInference, which provides the novel δ-token fairness guarantee: for a well-behaved client, if a token is generated in d time units in isolation, it will be generated within d + δ time units in multi-tenant execution, providing strong latency isolation guarantees for LLM serving. To achieve this, FairInference addresses a key challenge of LLM serving: bounding delays from sharing GPU resources without support for fine-grained scheduling or resource allocation. In FairInference, the scheduler enforces per-token deadlines, while bounding the delays from GPU compute sharing and accounting for the additional delays introduced by the shared KV caching in GPU memory. We show that FairInference effectively bounds token-level latency spikes for well-behaved clients and improves overall throughput compared to state-of-the-art LLM serving systems.

cs.DC↗

A Lightweight MPC Bidding Framework for Brand Auction Ads

Brand advertising plays a critical role in building long-term consumer awareness and loyalty, making it a key objective for advertisers across digital platforms. Although real-time bidding has been extensively studied, there is limited literature on algorithms specifically tailored for brand auction ads that fully leverage their unique characteristics. In this paper, we propose a lightweight Model Predictive Control (MPC) framework designed for brand advertising campaigns, exploiting the inherent attributes of brand ads -- such as stable user engagement patterns and fast feedback loops -- to simplify modeling and improve efficiency. Our approach utilizes online isotonic regression to construct monotonic bid-to-spend and bid-to-conversion models directly from streaming data, eliminating the need for complex machine learning models. The algorithm operates fully online with low computational overhead, making it highly practical for real-world deployment. Simulation results demonstrate that our approach significantly improves spend efficiency and cost control compared to baseline strategies, providing a scalable and easily implementable solution for modern brand advertising platforms.

cs.GT↗

Real-Time Generation of Streamable Talking Portrait Video with Reference-Guided Deep Compression VAEs

Video diffusion models have significantly advanced portrait video generation, yet their high computational demands limit their use in interactive applications. This work presents a framework for streamable talking portrait video generation conditioned on speech audio and reference images. Designed meticulously for streaming scenarios, it features a causal video VAE for deep latent compression and an autoregressive latent denoising model. Our causal VAE integrates a variable number of reference images as guidance, allowing the network to focus on dynamic information rather than static appearance, thereby enhancing compression efficacy and reconstruction quality. Additionally, we extend the residual auto-encoding paradigm to improve spatial-temporal causality handling in our VAE. The generator is based on a Rectified Flow Transformer architecture and produces video latents in a blockwise auto-regressive manner. Our method enables the real-time generation of high-quality talking portrait videos, achieving speeds significantly faster than baseline models. Furthermore, comprehensive experiments demonstrate that it is on par with or even outperforms these large models in realism, vividness, and video quality.

cs.CV↗

PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation

Augmenting large language models (LLMs) with retrieved web text has become a dominant paradigm, yet the web is not natively textual: existing systems depend on complex parsing pipelines that linearize HTML and discard layout, visual structure, and formatting. We introduce PixelRAG, a new retrieval-augmented method that represents websites in their native visual form and performs retrieval and reading entirely in pixel space, enabling an end-to-end architecture that eliminates text abstraction. PixelRAG is, to our knowledge, the first pipeline to operate over a full Wikipedia corpus in this form, scaling to a datastore of 30 million screenshot images with an efficient visual retrieval index. Built on an existing visual embedding model (i.e., Qwen3-VL-Embedding), PixelRAG further fine-tunes this model on screenshot data with carefully curated contrastive training data. Retrieved screenshots are then fed directly as pixel inputs to a VLM, without intermediate text conversion. PixelRAG consistently outperforms both no-retrieval and text-based RAG baselines, most surprisingly on widely studied text-centric tasks such as NQ and SimpleQA. It also achieves strong gains on multimodal open-domain QA (e.g., MMSearch), benchmarks over noisy news corpora (e.g., LiveVQA), and agentic benchmarks (e.g., MoNaCo), improving accuracy by up to 18.1% over text-based baselines. Finally, pixel representations enable a new efficiency lever for RAG through image compression, achieving up to 3x token cost reduction at lower resolutions while maintaining accuracy. Our results challenge the necessity of text representations in web retrieval, suggesting that web RAG can operate directly in the web's native visual form while improving both performance and efficiency.

cs.IR↗

FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale

Many real-world coding challenges are open-ended and admit no known optimal solution. Yet, recent progress in LLM coding has focused on well-defined tasks such as feature implementation, bug fixing, and competitive programming. Open-ended coding remains a weak spot for LLMs, largely because open-ended training problems are scarce and expensive to construct. Our goal is to synthesize open-ended coding problems at scale to train stronger LLM coders. We introduce FrontierSmith, an automated system for iteratively evolving open-ended problems from existing closed-ended coding tasks. Starting from competitive programming problems, FrontierSmith generates candidate open-ended variants by changing the problems'goals, restricting outputs, and generalizing inputs. It then uses a quantitative idea divergence metric to select problems that elicit genuinely diverse approaches from different solvers. Agents then generate test cases and verifiers for the surviving candidates. On two open-ended coding benchmarks, training on our synthesized data yields substantial gains over the base models: Qwen3.5-9B improves by +8.82 score on FrontierCS and +306.36 (Elo-rating-based performance) on ALE-bench; Qwen3.5-27B improves by +12.12 and +309.12, respectively. The synthesized problems also make agents take more turns and use more tokens, similar to human-curated ones, suggesting that closed-ended seeds can be a practical starting point for long-horizon coding data.

cs.LG↗

T-RBFT: A Scalable and Efficient Byzantine Consensus Based on Trusted Execution Environment for Consortium Blockchain

With the continuous expansion of blockchain application scenarios, consortium chains have raised higher performance and security requirements for consensus mechanisms. Unlike public blockchains, consortium chains typically implement an admission mechanism that restricts participation to trusted entities, ensuring that most replicas are honest and the number of faulty nodes remains small under normal circumstances. In such settings, conventional Byzantine Fault Tolerant (BFT) protocols, which are designed for worst-case adversarial scenarios, incur excessive message exchanges and computational overhead, thereby limiting performance and scalability. To address this issue, this paper proposes T-RBFT, a two-layer consensus mechanism inspired by network sharding and enhanced by the trusted execution environment (TEE). In T-RBFT, consensus nodes are first dynamically grouped based on their runtime characteristics. Then, inter-group consensus is achieved through a TEE-assisted BFT protocol, while each group internally reaches agreement using an improved Raft-based mechanism. Experimental evaluation shows that T-RBFT reduces communication overhead and latency, and achieves higher throughput compared to existing two-layer consensus protocols, providing a scalable and communication-efficient consensus protocol for permissioned blockchain networks.

cs.DC↗

Hardness of Range Avoidance and Proof Complexity Generators from Demi-Bits

Given a circuit $G: \{0, 1\}^n \to \{0, 1\}^m$ with $m > n$, the *range avoidance* problem ($\text{Avoid}$) asks to output a string $y\in \{0, 1\}^m$ that is not in the range of $G$. Besides its profound connection to circuit complexity and explicit construction problems, this problem is also related to the existence of *proof complexity generators* -- circuits $G: \{0, 1\}^n \to \{0, 1\}^m$ where $m > n$ but for every $y\in \{0, 1\}^m$, it is infeasible to prove the statement "$y\not\in\mathrm{Range}(G)$" in a given propositional proof system. This paper connects these two problems with the existence of *demi-bits generators*, a fundamental cryptographic primitive against nondeterministic adversaries introduced by Rudich (RANDOM '97). $\bullet$ We show that the existence of demi-bits generators implies $\text{Avoid}$ is hard for nondeterministic algorithms. This resolves an open problem raised by Chen and Li (STOC '24). Furthermore, assuming the demi-hardness of certain LPN-style generators or Goldreich' PRG, we prove the hardness of $\text{Avoid}$ even when the instances are constant-degree polynomials over $\mathbb{F}_2$. $\bullet$ We show that the dual weak pigeonhole principle is unprovable in Cook's theory $\mathsf{PV}_1$ under the existence of demi-bits generators secure against $\mathbf{AM}$, thereby separating Jerabek's theory $\mathsf{APC}_1$ from $\mathsf{PV}_1$. $\bullet$ We transform demi-bits generators to proof complexity generators that are *pseudo-surjective* with nearly optimal parameters. Our constructions build on the recent breakthroughs on the hardness of $\text{Avoid}$ by Ilango, Li, and Williams (STOC '23) and Chen and Li (STOC '24). We use *randomness extractors* to significantly simplify the construction and the proof.

cs.CC↗

FrontierCS: Evolving Challenges for Evolving Intelligence

We introduce FrontierCS, a benchmark of 156 open-ended problems across diverse areas of computer science, designed and reviewed by experts, including CS PhDs and top-tier competitive programming participants and problem setters. Unlike existing benchmarks that focus on tasks with known optimal solutions, FrontierCS targets problems where the optimal solution is unknown, but the quality of a solution can be objectively evaluated. Models solve these tasks by implementing executable programs rather than outputting a direct answer. FrontierCS includes algorithmic problems, which are often NP-hard variants of competitive programming problems with objective partial scoring, and research problems with the same property. For each problem we provide an expert reference solution and an automatic evaluator. Combining open-ended design, measurable progress, and expert curation, FrontierCS provides a benchmark at the frontier of computer-science difficulty. Empirically, we find that frontier reasoning models still lag far behind human experts on both the algorithmic and research tracks, that increasing reasoning budgets alone does not close this gap, and that models often over-optimize for generating merely workable code instead of discovering high-quality algorithms and system designs.

cs.LG↗

DS SERVE: A Framework for Efficient and Scalable Neural Retrieval

We present DS-Serve, a framework that transforms large-scale text datasets, comprising half a trillion tokens, into a high-performance neural retrieval system. DS-Serve offers both a web interface and API endpoints, achieving low latency with modest memory overhead on a single node. The framework also supports inference-time trade-offs between latency, accuracy, and result diversity. We anticipate that DS-Serve will be broadly useful for a range of applications, including large-scale retrieval-augmented generation (RAG), training data attribution, training search agents, and beyond.

cs.IR↗

LEANN: A Low-Storage Vector Index

Embedding-based vector search underpins many important applications, such as recommendation and retrieval-augmented generation (RAG). It relies on vector indices to enable efficient search. However, these indices require storing high-dimensional embeddings and large index metadata, whose total size can be several times larger than the original data (e.g., text chunks). Such high storage overhead makes it difficult, or even impractical, to deploy vector search on personal devices or large-scale datasets. To tackle this problem, we propose LEANN, a storage-efficient index for vector search that recomputes embeddings on the fly instead of storing them, and compresses state-of-the-art proximity graph indices while preserving search accuracy. LEANN delivers high-quality vector search while using only a fraction of the storage (e.g., 5% of the original data) and supporting storage-efficient index construction and updates. On real-world benchmarks, LEANN reduces index size by up to 50x compared with conventional indices, while maintaining SOTA accuracy and comparable latency for RAG applications.

cs.DB↗

MuseGNN: Forming Scalable, Convergent GNN Layers that Minimize a Sampling-Based Energy

Among the many variants of graph neural network (GNN) architectures capable of modeling data with cross-instance relations, an important subclass involves layers designed such that the forward pass iteratively reduces a graph-regularized energy function of interest. In this way, node embeddings produced at the output layer dually serve as both predictive features for solving downstream tasks (e.g., node classification) and energy function minimizers that inherit transparent, exploitable inductive biases and interpretability. However, scaling GNN architectures constructed in this way remains challenging, in part because the convergence of the forward pass may involve models with considerable depth. To tackle this limitation, we propose a sampling-based energy function and scalable GNN layers that iteratively reduce it, guided by convergence guarantees in certain settings. We also instantiate a full GNN architecture based on these designs, and the model achieves competitive accuracy and scalability when applied to the largest publicly-available node classification benchmark exceeding 1TB in size. Our source code is available at https://github.com/haitian-jiang/MuseGNN.

cs.LG↗

Autellix: An Efficient Serving Engine for LLM Agents as General Programs

Large language model (LLM) applications are evolving beyond simple chatbots into dynamic, general-purpose agentic programs, which scale LLM calls and output tokens to help AI agents reason, explore, and solve complex tasks. However, existing LLM serving systems ignore dependencies between programs and calls, missing significant opportunities for optimization. Our analysis reveals that programs submitted to LLM serving engines experience long cumulative wait times, primarily due to head-of-line blocking at both the individual LLM request and the program. To address this, we introduce Autellix, an LLM serving system that treats programs as first-class citizens to minimize their end-to-end latencies. Autellix intercepts LLM calls submitted by programs, enriching schedulers with program-level context. We propose two scheduling algorithms-for single-threaded and distributed programs-that preempt and prioritize LLM calls based on their programs' previously completed calls. Our evaluation demonstrates that across diverse LLMs and agentic workloads, Autellix improves throughput of programs by 4-15x at the same latency compared to state-of-the-art systems, such as vLLM.

cs.LG↗