Searcharxiv⌕ Search

arXiv subjects

Yicheng Wang

Publications and source records attributed to Yicheng Wang.

At least 19 recordsLinked to original sources

JoyFrontBench: Evaluating Front-End Coding Agents through Goal-Oriented Acceptance Contracts

Front-end coding agents are expected to produce applications that fulfill user goals, yet high task scores can coexist with failures in essential workflows. Evaluating these agents therefore requires an explicit distinction between partial implementation and application success. We introduce JoyFrontBench, a benchmark that makes this distinction through goal-oriented acceptance contracts. Each contract identifies core tasks and mandatory criteria before generation: a core task passes only when all mandatory criteria are satisfied and a shared completion threshold is met, while application success requires every core task to pass on the same operational artifact within a single run. Derived from practitioner requirements, JoyFrontBench comprises 100 front-end development problems, 764 tasks, and 2,596 atomic criteria, with reviewed annotations linking user goals to core tasks and evidence requirements. The evaluation framework combines evidence-grounded judgments with deterministic record auditing and fixed scoring rules, preserving graded task assessment alongside explicit core-task and application-level acceptance. We evaluate five coding models under two agent harnesses, OpenHands and OpenCode, on the 70-problem test set; the remaining 30 problems form a public set for validation and diagnostic analyses. Under OpenHands, the mean Average Task Score (ATS) across models is 87.69%, while the mean Application Success Rate (ASR) is 51.52%. Further analyses reveal changes in model rankings across harnesses and variation in application success across repeated runs. These findings highlight the limitations of aggregate task scores in characterizing application success and motivate explicit assessment of core-goal fulfillment alongside graded task performance.

cs.SE↗

Cross-Day Compatibility of Daily Item Embeddings: An Empirical Study of Alignment, Scale Management, and CTR Adaptation

Daily item co-occurrence embeddings are paired with next-day behavior without future-behavior leakage. Multi-day click-through rate (CTR) training therefore mixes versions whose directions and norms may differ for the same item. We compare Word2Vec and an item-graph XSimGCL adaptation across initialization, inheritance, post-processing, training constraints, and downstream augmentation. The inventory contains 48 configurations, including 32 common-protocol CTR evaluations. Observed metrics are reported as relative percentage changes from the corresponding Word2Vec baseline (0%), with raw values and exact dataset sizes withheld. Three findings emerge. First, multiple upstream interventions improve compatibility and CTR. ID-init and orthogonal alignment increase cosine by 51.2% and 93.4%, and GAUC by 2.2% and 2.6%, relative to ordinary Word2Vec. Retrieval and geometry remain insufficient downstream surrogates. Second, warm starts perform strongly but accumulate scale. The day-matched norm increase over Word2Vec reaches 102.5% after eight days; daily global rescaling limits it to 8.2% and attains the highest observed GAUC increase, 4.0%. Third, CTR learning through a dimensionality-reducing encoder further adapts to residual drift. Against dimension-matched random-projection, SVD, and untrained-network controls, trained encoders attain the highest cross-day cosine for four of six upstream methods. The other two rank third and second in train-to-test cosine. These industrial results support complementary roles for upstream stabilization, inherited-scale management, and downstream task learning.

cs.IR↗

AudioFM: Towards General Purpose Audio Representation Learning

Learning audio representations that effectively bridge holistic global soundscapes with fine-grained acoustic moments is a long-standing goal in self-supervised learning. Recording-level contrastive objectives encourage agreement between pooled audio representations, but do not explicitly supervise local time-frequency representations. We introduce AudioFM, a self-supervised framework that combines recording-level contrast with patch-wise masked latent discrimination. Using visible acoustic context, a student predicts representations of masked patches and identifies their corresponding full-view teacher targets among candidates from other recordings. We further incorporate weaker supervision from eligible temporal neighbors to encourage local temporal consistency. Extensive evaluations show that AudioFM achieves state-of-the-art performance across diverse audio domains in linear probing, fine-tuning, and zero-shot transfer, demonstrating superior event detection, robust scene-level understanding, and strong domain generalizability.

cs.SD↗

ClueFlow: Learning Long-Horizon Search through Attributed Research-State Transitions

Long-horizon search requires agents to acquire relevant evidence and integrate it into an evolving research state. Existing methods maintain research states to guide exploration or assign turn-level progress signals, yet grounding these signals in explicit, attributable state changes remains underexplored. We introduce ClueFlow, which learns long-horizon search from attributed research-state transitions. ClueFlow constructs Source-State-Final attribution chains that connect evidence-acquiring browser actions, Evidence and Candidate updates, and the final answer supported by those updates. During search, ClueFlow records typed state changes and potentially supporting browser actions, then applies gold-conditioned semantic validation and source tracing after execution. We use this attribution to guide both supervised fine-tuning and reinforcement learning. During SFT, attribution selects which decisions receive supervision with their original interaction contexts. During Turn-PPO, the same attribution augments outcome-based turn advantages with credit assigned along the Source-State-Final chains. Transition-level validation also recovers useful supervision from gold-consistent state changes in trajectories with incorrect final answers. We evaluate ClueFlow across two Qwen backbones and four search benchmarks spanning live-web and fixed-corpus settings, observing consistent gains from SFT and further improvements from PPO. With Qwen3.5-35B-A3B, ClueFlow achieves 60.0% accuracy on BrowseComp and 72.0% on BrowseComp-Plus. In controlled BrowseComp-Plus experiments, attribution-guided SFT matches all-turn exact-context SFT using about 30% of the training tokens, while attribution-guided PPO outperforms outcome-only PPO by 4.1 pp. These results demonstrate that state-transition attribution enables efficient supervision and effective credit assignment for long-horizon search agents.

cs.AI↗

Watt-level single-cavity dual-comb laser at 2.1 $μ$m wavelength

Compact ultrafast laser combs around the 2 $μ$m atmospheric transmission window are of growing interest for direct spectroscopy and as drivers for nonlinear conversion to otherwise hard-to-reach spectral regions. Dual-comb systems in this region remain very scarce: existing implementations rely mostly on nonlinear conversion, resulting in complex systems with poor conversion efficiency and overall low average and peak power. In this paper, we report on a watt-level dual-comb laser directly emitting at 2.1 $μ$m. The system is based on a single-cavity, soft-aperture Kerr-lens mode-locked (KLM) laser using a broadband holmium (Ho)-doped CaAlGdO$_4$ (CALGO) gain crystal. The laser delivers a total output power exceeding 2 W (1 W per comb) with a sub-100-fs pulse duration (50 nm bandwidth) at a repetition rate of 96.5 MHz, corresponding to a pulse energy of 10 nJ and a peak power of 100 kW per comb. To assess the source's potential for future spectroscopic applications, we characterize the free-running dual comb in detail for frequency stability, phase noise, and relative intensity noise (RIN), and benchmark the achievable spectral resolution by measuring spectral modulations imposed by Fabry-Pérot etalons. Our system combines the compactness, high efficiency, and passive mutual coherence of a single-cavity oscillator with watt-level average power at 2.1 $μ$m, making it a promising source for environmental sensing in the atmospheric transmission window, or as a driver for nonlinear frequency conversion, for example, to the MIR and THz regions.

physics.optics↗

Avoiding What Your Teacher Already Knows: Entity Familiarity for Deep-Search Distillation

Deep-search agents are commonly trained by distilling tool-use trajectories from stronger teacher models. However, structurally complex questions may remain relatively easy for teachers familiar with key entities in the underlying entity graph, resulting in lower search demand and weaker supervision. We refer to this teacher-specific variation in question difficulty as teacher-relative difficulty and investigate entity familiarity as an early predictor. We first introduce Evidence-Verified Entity Familiarity (EVF), which estimates familiarity by verifying a teacher-generated factual brief against online evidence. Across multiple teachers, EVF is significantly correlated with teacher-relative difficulty with AUCs of 0.60--0.70. However, its high cost makes it impractical for constructing entity graphs at scale. To address this limitation, we propose Self-Reported Entity Familiarity (SEF), a lightweight gating signal that directly asks a teacher whether it is familiar with an entity before admitting that entity into the graph. Across multiple teachers, trajectories from SEF-gated synthesis exhibit greater observed search demand and yield higher downstream student accuracy than those from ungated synthesis. Because familiarity is teacher specific, we further study cross-teacher reuse. We screen graphs collected for other teachers using Kimi-K3's SEF judgments and use the retained graphs to produce 5,411 additional trajectories. Combined with 2,200 trajectories from newly constructed, SEF-gated graphs, these data improve Qwen3.8-27B's accuracy from 44.10% to 66.41% on a BrowseComp subset, and achieving state-of-the-art performance among models of comparable size on BrowseComp, SEAL-0, HLE-text, and xbench. Overall, SEF provides a low-cost way to synthesize more informative supervision for deep-search distillation.

cs.CL↗

BVB: Benchmarking Agentic Video Understanding via Programmatic Reconstruction in Blender

Multimodal agents can create complex videos in software such as Blender by writing code instead of using diffusion models. Yet video understanding benchmarks still evaluate models mainly through question answering. If an agent truly understands a video, it can reconstruct it programmatically. We introduce BVB, Blender-VideoBench, a benchmark that tests this ability by asking agents to reconstruct real-world videos as animated Blender scenes. To ensure fair comparison, each agent programs the reconstruction through a lightweight harness, Mini-BVB, in an identical sandbox under a shared cost limit. The benchmark renders each reconstruction from its animated camera and evaluates it on two axes: (1) Dual VQA measures how many spatiotemporal facts the reconstruction preserves. (2) Latent Similarity measures how closely the reconstruction matches the source video perceptually. Our overall score, a square-root mean, favors balanced performance. We evaluate 51 configurations from 10 model families and analyze semantic retention, perceptual similarity, reasoning effort, and cost. The best model reaches 88.6 Latent Similarity but retains only 53.7% of the spatiotemporal facts from the source video. Additional reasoning improves perceptual similarity but does not close this gap. In a blind study with 15 raters and five configurations, Latent Similarity correlates strongly with human preference. These results show that programmatic reconstruction is a viable test of agentic video understanding, and that semantic retention remains the main challenge.

cs.CV↗

DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale

Large-scale agentic training and evaluation with large language models (LLMs) rely on isolated, stateful execution environments in which models inspect repositories, invoke tools, execute commands, and interact with task-specific services. These workloads create sandboxes in large bursts, span heterogeneous functionality and isolation requirements, retain state across long interactions, and draw from large image corpora with limited reuse. Supporting them therefore requires an elastic execution platform rather than a single sandbox runtime. This report presents DeepSeek Elastic Compute (DSec), a production sandbox platform that exposes FnCall, container, microVM, and full-VM sandbox backends through a unified SDK. DSec coordinates placement and lifecycle management across the cluster, composes environments from independently versioned layers, combines memory sharing, reclamation, and CPU scheduling for high-density execution, and loads image data on demand from Fire-Flyer File System (3FS), a cluster-wide distributed filesystem. DSec is co-designed with the reinforcement learning (RL) framework, decouples stateful rollout execution from preemptible GPU training, coordinates sandbox lifecycle with training to preserve rollout state while reclaiming idle resources, and mitigates agent misbehavior such as reward hacking. A single production-scale unit of DSec spans around 160 nodes, serving about 3 million sandboxes per day; in production, it supports over 380,000 concurrent sandboxes and sustains over 5,000 sandbox creations per second. Our evaluation and deployment experience show that these mechanisms reduce environment setup and image-distribution overhead, improve memory efficiency, and preserve latency-sensitive performance under high-density overcommit.

cs.DC↗

Intracavity THz generation using a thin lithium niobate plate in a compact Kerr-lens mode-locked Yb:CALGO bulk oscillator

We demonstrate intracavity terahertz (THz) generation via optical rectification in a 50-m-thick lithium niobate crystal placed inside a compact diode-pumped Kerr-lens mode-locked (KLM) Yb:CALGO bulk oscillator. The oscillator operates at a repetition rate of 85 MHz and delivers 83-fs pulses with up to 71 W of average intracavity power, obtained with only 21.4 W of low-cost multimode diode pump power. We generate single-cycle THz pulses with a spectrum extending up to 3 THz, detected by electro-optic sampling with 60 dB dynamic range within 156 s of measurement time (313 averaged traces) and up to 120 W of THz average power. This work combines the high damage threshold, power-handling capability, and cost-effectiveness of thin LN plates with simplicity, compactness, and low-cost multimode diode-pumped solid-state bulk lasers, offering an attractive alternative for high-repetition-rate THz time-domain spectroscopy systems.

physics.optics↗

An Atlas and Design Rules for Single- and Dual-Atom Alloys

A long-standing goal across heterogeneous catalysis, materials science, and condensed matter physics is to design alloys with prescribed local atomic arrangements. Recent experiments show that dilute trimetallic alloys unlock chemistries inaccessible to bimetallics, but realizing this potential requires knowing which dopant structures form across an enormous compositional space. Using density functional theory screening, we construct an atlas spanning transition metal single- and dual-atom alloys in Cu and Ag surfaces, which we validate by scanning tunneling microscopy. The stability follows an electron count: dopants pair most strongly when their combined d-electron count approaches ten. Host metal and surface facet can be used to tune the resulting active site motifs, while size mismatch and spin explain variations around this trend. We further introduce a reactor-anchor concept, in which one dopant anchors a second, otherwise bulk-segregating dopant at the surface. Together, these results establish design principles for engineering alloys with targeted active site structures.

cond-mat.mtrl-sci↗

Redshift evolution of the Hubble constant: Constraints and new insights from an interacting dark energy model

We develop a modified interacting dark energy (IDE) model to study the redshift evolution of the Hubble constant ($H_0$), in light of the Hubble tension. In this framework, the energy exchange between dark energy and dark matter induces a redshift dependence of $H_0$. We evaluate the model against a comprehensive suite of observations, including baryon acoustic oscillations (BAO) from DESI DR2 and SDSS, cosmic chronometers, type Ia supernovae from the Pantheon sample, and Planck CMB distance priors. Analysis of late-Universe data yields $α= 0.0107^{+0.0032}_{-0.011}$, with the best-fit value on the order of $10^{-2}$, revealing a decreasing trend of $H_0$ with redshift. This supports a power-law evolution beyond $Λ$CDM. Incorporating CMB data further tightens the constraint to the order of $10^{-5}$, which we attribute to the suppression of dark-sector interactions at high redshifts, a consequence of the strong baryon--photon coupling. These results indicate that the IDE framework provides a theoretically consistent and observationally viable mechanism for describing the redshift evolution of $H_0$, offering a promising avenue toward alleviating the Hubble tension.

astro-ph.CO↗

Cosmological Constraints on the DGP Model in light of DESI DR2 2025 Data

We present updated constraints on both flat and non-flat Dvali-Gabadadze-Porrati (DGP) cosmological models using the latest baryon acoustic oscillation (BAO) measurements from the Dark Energy Spectroscopic Instrument Data Release 2 (DESI DR2), in combination with cosmic chronometer (CC), Type Ia supernova (SNIa), and cosmic microwave background (CMB) distance priors. For the non-flat DGP model, we obtain $H_0 = 64.05 \pm 0.27\, \rm{kms^{-1}Mpc^{-1}}$, $Ω_m = 0.3264 \pm 0.0043$, and $Ω_k = 0.0088 \pm 0.0016$, corresponding to a transition redshift $z_t \sim 0.41$. For the flat case, the constraints are $H_0 = 63.28 \pm 0.25\, \rm{kms^{-1}Mpc^{-1}}$ and $Ω_m = 0.3303 \pm 0.0036$. In both scenarios, the inferred Hubble constant is significantly lower than the Planck $\mathrmΛ$CDM value, indicating that the DGP framework does not alleviate the Hubble tension. Current observations strongly disfavor the DGP framework, primarily due to its inability to simultaneously accommodate DESI BAO and CMB constraints.By incorporating the latest high-precision DESI observations within a unified analysis framework, this work provides updated and more stringent limits on the DGP scenario, offering a consolidated assessment of its viability in the context of current cosmological data.

astro-ph.CO↗

CollabSkill: Evaluating Human-Agent Collaboration On Real-World Tasks

AI agents are reshaping the workspace, leading to drastic change of how humans work. Despite the considerable potential of human-agent collaboration both in preserving human agency and generating economic value, this paradigm remains largely absent from occupational task evaluation, hindered by the difficulty of gathering real human data and accounting for inter-human variability. We introduce CollabSkill, a framework for evaluating human-agent collaboration on real-world occupational tasks. CollabSkill pairs real human workers with AI agents on tasks matched to their occupational background, collecting data that capture the complexity of economically valuable tasks and the usage patterns of real workers. To account for inter-human variability, CollabSkill employs a Bayesian skill rating system to disentangle and quantify the skill contributions of both humans and AI agents. Drawing on over 1,500 prompts from 386 working sessions contributed by 93 human workers, our analysis yields insights on two fronts: on the agent side, rankings on CollabSkill diverge meaningfully from those of existing fully autonomous benchmarks where Codex leads, with Claude Code ranking first; on the human side, CollabSkill reveals that practical experience emerges as the primary driver of collaboration skill, with hands-on collaboration meaningfully shifting workers' AI literacy. Together, we hope CollabSkill enables the community to invest in systematic evaluation of human-agent collaboration and spurs development efforts aimed at building AI agents that genuinely augment human workers.

cs.HC↗

Diagnosing and Mitigating Retrieval Bottlenecks in LLM-Based Cold-Start Recommendation

Large language models (LLMs) are increasingly used as rerankers in recommender systems, with the expectation that semantic understanding will help in cold-start and long-tail regimes. We test this assumption with a five-domain benchmark that explicitly separates reranking quality from retrieval coverage. In a positive-controlled regime where the gold item is guaranteed present, calibrated LLM rerankers fail to consistently outperform strong collaborative and content baselines under natural traffic, and within-family scaling from Qwen3-8B to Qwen3-32B narrows but does not close the gap on most domains. In a retrieval-realistic regime where the gold item is not injected, the bottleneck is more severe: standard single retrievers place the gold item in a 200-item pool only 4.6-22.9% of the time, largely because 32-91% of cold-start targets are brand-new items with no training interactions. We introduce LHF, a validation-trained learned hybrid fusion layer over a multi-retriever union pool, as a retrieval-side realizability baseline. LHF is the only combiner we test that beats every single retriever on all five domains and recovers 17-61% of oracle coverage headroom on content-rich domains, but only 5-7% on collaboratively strong domains. End-to-end experiments reveal the remaining mismatch: learned non-LLM ranking exploits the LHF pool, while prompt-level LLM reranking often degrades it. LLMs exhibit pockets of semantic cold-start advantage, especially in text-rich domains when the item is already present, but this advantage is largely unreachable in current retrieve-then-rerank pipelines. We release the benchmark protocol, splits, prompts, evaluation tooling, and archived reproducibility artifacts: data at https://doi.org/10.5281/zenodo.20991039 and code at https://doi.org/10.5281/zenodo.20993306.

cs.IR↗

Know Before You Fetch: Calibrated Retrieval-Budget Allocation for Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) typically retrieves a fixed number of passages for every query. This is wasteful when the reader already knows the answer, and it can be harmful when irrelevant or partially relevant passages distract the reader. We formulate adaptive RAG as calibrated retrieval-budget allocation: given a query, decide whether to answer closed-book, retrieve a compact context (k=1), retrieve a full context (k=5), or abstain. The contribution is a probability interface rather than a new raw uncertainty signal. We calibrate sequence log-probability and prefix-logit uncertainty signals into probabilities of correctness, then use these probabilities for graded context selection, selective abstention, and explicit latency/token trade-offs. Across core QA experiments on TriviaQA, Natural Questions, and MS MARCO, with auxiliary PopQA motivation and Qwen/Llama family checks, diagnostic out-of-fold calibration improves probability quality dramatically: for sequence log-probability, ECE drops from 0.275 to 0.062 on TriviaQA, 0.643 to 0.009 on NQ, and 0.711 to 0.031 on MS MARCO. Graded retrieval improves full-context and passage-budget frontiers for both our signal and TARG-style prefix entropy/margin, while retrieval-call AUC remains essentially tied with binary gating because k=1 is still a retrieval call. Held-out train/validation/test threshold experiments report deployable operating points. At matched-accuracy frontier operating points, a measured cost model reveals that gating is not universally faster: it increases latency by about 27% on Qwen3-8B but saves about 8% on Qwen3-32B. These results support a nuanced view of adaptive RAG: calibrated confidence is best understood as a reusable interface for allocating retrieval budget under task and system constraints.

cs.IR↗

Cosmological constraints on the big bang quantum cosmology model

The big bang quantum cosmology model introduces the trace $J$ of the Schouten tensor as a form of dynamic dark energy. Together with cold dark matter, these components form the so-called $J$CDM cosmology model, proposed by M.H.P.M. van Putten (J. High Energy Astrophys., 45, 2025, 194), which offers a potential resolution to the Hubble tension. We derive the constraints on the $J$CDM cosmology model, utilizing early- and late-time cosmological data including cosmic microwave background (CMB), baryon acoustic oscillations (BAO) released by the Dark Energy Spectroscopic Instrument (DESI), cosmic chronometers (CC), and type Ia supernovae (SNIa). For a flat universe, the $J$CDM model yields \( H_0 = 66.95 \pm 0.51 \, \rm{km~s^{-1}~Mpc^{-1}} \) and \( Ω_m = 0.3419 \pm 0.0065 \), results that are consistent with early-universe observations but exhibit a higher \( Ω_m \) compared to the $Λ$CDM model. In the case of a non-flat universe, $J$CDM favors a slightly curved geometry with \( Ω_k = 0.0154 \pm 0.0027 \), leading to \( H_0 = 69.13 \pm 0.56 \, \rm {km~s^{-1}~Mpc^{-1}} \) and \( Ω_m = 0.3477 \pm 0.0074 \). The increase in \( H_0 \) in the non-flat scenario suggests a geometric degeneracy between spatial curvature and \( H_0 \). We also investigate the internal inconsistencies present in DESI data and evaluate their impacts on cosmological parameter constraints. Our analysis shows that while the $J$CDM model, which is constructed from first principles without free parameters beyond those of $Λ$CDM, agrees excellently with late-time cosmology, it struggles to simultaneously match early-universe observations in a fully self-consistent manner.

astro-ph.CO↗

Optimization of Laser Irradiation Uniformity for the Double-Cone Ignition Scheme with MULTI-3D simulations

The double-cone ignition (DCI) scheme holds a promising perspective for laser driven fusion energy and astrophysics. However, optimizing the laser irradiation uniformity under the constraints of limited laser beams and a given cone angle remains to be explored. We utilized the three-dimensional radiation hydrodynamics program MULTI-3D to simulate the interaction process between the laser and plasma shell. By employing Bayesian optimization for the pointing position of the incident laser beams, we achieved a laser irradiation scheme with nonuniformity less than 5%. This study can provide references for experiments and offer valuable insights for other laser fusion schemes.

physics.plasm-ph↗

Qianfan-OCR: A Unified End-to-End Model for Document Intelligence

We present Qianfan-OCR, a 4B-parameter end-to-end vision-language model that unifies document parsing, layout analysis, and document understanding within a single architecture. It performs direct image-to-Markdown conversion and supports diverse prompt-driven tasks including table extraction, chart understanding, document QA, and key information extraction. To address the loss of explicit layout analysis in end-to-end OCR, we propose Layout-as-Thought, an optional thinking phase triggered by special think tokens that generates structured layout representations -- bounding boxes, element types, and reading order -- before producing final outputs, recovering layout grounding capabilities while improving accuracy on complex layouts. Qianfan-OCR ranks first among end-to-end models on OmniDocBench v1.5 (93.12) and OlmOCR Bench (79.8), achieves competitive results on OCRBench, CCOCR, DocVQA, and ChartQA against general VLMs of comparable scale, and attains the highest average score on public key information extraction benchmarks, surpassing Gemini-3.1-Pro, Seed-2.0, and Qwen3-VL-235B. The model is publicly accessible via the Baidu AI Cloud Qianfan platform.

cs.CV↗