arXiv subjects

Chenghao Zhu

Publications and source records attributed to Chenghao Zhu.

At least 19 recordsLinked to original sources

arXiv2026

SDSS+JWST Census of Stellar and Nebular Dust Attenuation at $z \sim 0$-7: Mass Dependence and Redshift Evolution

We present the demography of dust attenuation, including its mass dependence and redshift evolution, using spectroscopic samples of 34,182 SDSS galaxies at $z\sim0.1$ and 863 JWST/JADES galaxies at $z\sim1.5$--$7$. We find that, on average, ${\rm Hα}/{\rm Hβ}$ ratios are comparable to the Case B recombination value at $M_\ast \lesssim 10^9 M_\odot$, and increase beyond $M_\ast \sim 10^9 M_\odot$ both at $z\sim0.1$ and $1.5$--$7$. We derive the nebular attenuation $A_{\rm V, nebular}$ from Balmer decrements and the stellar attenuation $A_{\rm V, stellar}$ from rest-frame UV--optical spectra with supplementary \textit{GALEX} data, via comparisons with stellar-population models and multiple attenuation curves in a consistent manner across cosmic time. We find no significant redshift evolution of $A_{\rm V, nebular}$ and $A_{\rm V, stellar}$ at fixed $M_\ast$ over $z\sim0$--$7$, forming a universal extinction relation, and both rise from $0.2$--$0.4$ at $M_\ast \lesssim 10^9 M_\odot$ to $\sim1$ at $M_\ast \sim 10^{11} M_\odot$. Interestingly, at $M_\ast \gtrsim 10^9 M_\odot$, $A_{\rm V, nebular}$ rises more steeply than $A_{\rm V, stellar}$. This correlation holds within an uncertainty of $\sim\pm0.2$ for various combinations of attenuation curves (Calzetti, SMC, and Milky Way). These results indicate that $M_\ast \sim 10^9 M_\odot$ is a transition mass in dust attenuation, whose low-mass behavior reflects dust widely distributed by feedbacks. These mass-dependent extinction results address the long-standing issue of appropriate choice of the stellar-to-nebular color excess ratio, $f\equiv E(B-V)_{\rm stellar}/E(B-V)_{\rm nebular}=1.0$ or $0.44$, and suggest that galaxy $M_\ast$ determines $f$ from $\sim1.0$ to $\sim0.44$ across low- to high-mass galaxies.

astro-ph.GA↗

arXiv2026

LOGIGEN: Logic-Driven Generation of Verifiable Agentic Tasks

The evolution of Large Language Models (LLMs) from static instruction-followers to autonomous agents necessitates operating within complex, stateful environments to achieve precise state-transition objectives. However, this paradigm is bottlenecked by data scarcity, as existing tool-centric reverse-synthesis pipelines fail to capture the rigorous logic of real-world applications. We introduce \textbf{LOGIGEN}, a logic-driven framework that synthesizes verifiable training data based on three core pillars: \textbf{Hard-Compiled Policy Grounding}, \textbf{Logic-Driven Forward Synthesis}, and \textbf{Deterministic State Verification}. Specifically, a Triple-Agent Orchestration is employed: the \textbf{Architect} compiles natural-language policy into database constraints to enforce hard rules; the \textbf{Set Designer} initializes boundary-adjacent states to trigger critical policy conflicts; and the \textbf{Explorer} searches this environment to discover causal solution paths. This framework yields a dataset of 20,000 complex tasks across 8 domains, where validity is strictly guaranteed by checking exact state equivalence. Furthermore, we propose a verification-based training protocol where Supervised Fine-Tuning (SFT) on verifiable trajectories establishes compliance with hard-compiled policy, while Reinforcement Learning (RL) guided by dense state-rewards refines long-horizon goal achievement. On $τ^2$-Bench, LOGIGEN-32B(RL) achieves a \textbf{79.5\% success rate}, substantially outperforming the base model (40.7\%). These results demonstrate that logic-driven synthesis combined with verification-based training effectively constructs the causally valid trajectories needed for next-generation agents.

cs.AI↗

arXiv2026

QianfanHuijin Technical Report: A Novel Multi-Stage Training Paradigm for Finance Industrial LLMs

Domain-specific enhancement of Large Language Models (LLMs) within the financial context has long been a focal point of industrial application. While previous models such as BloombergGPT and Baichuan-Finance primarily focused on knowledge enhancement, the deepening complexity of financial services has driven a growing demand for models that possess not only domain knowledge but also robust financial reasoning and agentic capabilities. In this paper, we present QianfanHuijin, a financial domain LLM, and propose a generalizable multi-stage training paradigm for industrial model enhancement. Our approach begins with Continual Pre-training (CPT) on financial corpora to consolidate the knowledge base. This is followed by a fine-grained Post-training pipeline designed with increasing specificity: starting with Financial SFT, progressing to Finance Reasoning RL and Finance Agentic RL, and culminating in General RL aligned with real-world business scenarios. Empirical results demonstrate that QianfanHuijin achieves superior performance across various authoritative financial benchmarks. Furthermore, ablation studies confirm that the targeted Reasoning RL and Agentic RL stages yield significant gains in their respective capabilities. These findings validate our motivation and suggest that this fine-grained, progressive post-training methodology is poised to become a mainstream paradigm for various industrial-enhanced LLMs.

cs.CL↗

arXiv2026

Exact Clique Number Manipulation via Edge Interdiction

The Edge Interdiction Clique Problem (EICP) aims to remove at most $k$ edges from a graph so as to minimize the size of the largest clique in the remaining graph. This problem captures a fundamental question in graph manipulation: which edges are structurally critical for preserving large cliques? Such a problem is also motivated by practical applications including protein function maintenance and image matching. The EICP is computationally challenging and belongs to a complexity class beyond NP. Existing approaches rely on general mixed-integer bilevel programming solvers or reformulate the problem into a single-level mixed integer linear program. However, they are still not scalable when the graph size and interdiction budget $k$ grow. To overcome this, we investigate new mixed integer linear formulations, which recast the problem into a sequence of parameterized Edge Blocker Clique Problems (EBCP). This perspective decomposes the original problem into simpler subproblems and enables tighter modeling of clique-related inequalities. Furthermore, we propose a two-stage exact algorithm, \textsc{RLCM}, which first applies problem-specific reduction techniques to shrink the graph and then solves the reduced problem using a tailored branch-and-cut framework. Extensive computational experiments on maximum clique benchmark graphs, large real-world sparse networks, and random graphs demonstrate that \textsc{RLCM} consistently outperforms existing approaches.

cs.DS↗

arXiv2025

Towards Faithful and Controllable Personalization via Critique-Post-Edit Reinforcement Learning

Faithfully personalizing large language models (LLMs) to align with individual user preferences is a critical but challenging task. While supervised fine-tuning (SFT) quickly reaches a performance plateau, standard reinforcement learning from human feedback (RLHF) also struggles with the nuances of personalization. Scalar-based reward models are prone to reward hacking which leads to verbose and superficially personalized responses. To address these limitations, we propose Critique-Post-Edit, a robust reinforcement learning framework that enables more faithful and controllable personalization. Our framework integrates two key components: (1) a Personalized Generative Reward Model (GRM) that provides multi-dimensional scores and textual critiques to resist reward hacking, and (2) a Critique-Post-Edit mechanism where the policy model revises its own outputs based on these critiques for more targeted and efficient learning. Under a rigorous length-controlled evaluation, our method substantially outperforms standard PPO on personalization benchmarks. Personalized Qwen2.5-7B achieves an average 11\% win-rate improvement, and personalized Qwen2.5-14B model surpasses the performance of GPT-4.1. These results demonstrate a practical path to faithful, efficient, and controllable personalization.

cs.CL↗

arXiv2025

A Near-Real-Time Reduction-Based Algorithm for Coloring Massive Graphs

The graph coloring problem is a classical combinatorial optimization problem with important applications such as register allocation and task scheduling, and it has been extensively studied for decades. However, near-real-time algorithms that can deliver high-quality solutions for very large real-world graphs within a strict time frame remain relatively underexplored. In this paper, we try to bridge this gap by systematically investigating reduction rules that shrink the problem size while preserving optimality. For the first time, domination reduction, complement crown reduction, and independent set reduction are applied to large-scale instances. Building on these techniques, we propose RECOL, a reduction-based algorithm that alternates between fast estimation of lower and upper bounds, graph reductions, and heuristic coloring. We evaluate RECOL on a wide range of benchmark datasets, including SNAP, the Network Repository, DIMACS10, and DIMACS2. Experimental results show that RECOL consistently outperforms state-of-the-art algorithms on very large sparse graphs within one minute. Additional experiments further highlight the pivotal role of reduction techniques in achieving this performance.

cs.DS↗

arXiv2025

A High Incidence of Mid-infrared Variability in Local Ultraluminous Infrared Galaxies

We explore mid-infrared (MIR) variability in local ultraluminous infrared galaxies (ULIRGs; infrared luminsoity $L_{\rm IR}>10^{12}\ L_\odot$) utilizing the $\sim$11 years of photometry from the NEOWISE multi-epoch catalog of {\it Wide-field Infrared Survey Explorer} ({\it WISE}). We identify 30 variable ULIRGs with statistically significant MIR variability. The variability is observed on timescales of a few years, implying that the MIR-emitting regions are compact ($\lesssim 1$ pc). The difference between maximum and minimum $W2$ (4.6 ${\rm μ}$m) band luminosity ($ΔL_{\rm W2}$) of the 30 variable ULIRGs range from $ΔL_{W2}$ = $7\times10^{42}$ to $5\times 10^{44}$ erg s$^{-1}$. The $ΔL_{W2}$ of 25 variable ULIRGs out of 30 are greater than $ΔL_{W2}$ = $1\times10^{43}$ erg s$^{-1}$, surpassing the MIR luminosity {range} observed in known supernovae (SNe; $L_{\rm 3.6\ {\rm μm}}$ and $L_{\rm 4.5\ {\rm μm}}$ < 10$^{42.3}$ erg s$^{-1}$). Therefore, the MIR variabilities in these 25 ULIRGs are most likely driven by tidal disruption events (TDEs) or intrinsic changes in their active galactic nuclei (AGN) torus emission. Our sample includes hard X-ray detected AGNs (e.g., UGC 05101) and previously reported TDE candidates (IRAS F01004-2237, IRAS 05189-2524). All 25 also exhibit at least one AGN signature(s) besides the MIR variability, suggesting that even if the MIR variability originates from TDEs, the black holes responsible are likely AGNs. Our results suggest that MIR variability is an effective tool for detecting buried AGNs and highlights the intense nuclear activity in ULIRGs.

astro-ph.GA↗

arXiv2025

TalkVid: A Large-Scale Diversified Dataset for Audio-Driven Talking Head Synthesis

Audio-driven talking head synthesis has achieved remarkable photorealism, yet state-of-the-art (SOTA) models exhibit a critical failure: they lack generalization to the full spectrum of human diversity in ethnicity, language, and age groups. We argue that this generalization gap is a direct symptom of limitations in existing training data, which lack the necessary scale, quality, and diversity. To address this challenge, we introduce TalkVid, a new large-scale, high-quality, and diverse dataset containing 1244 hours of video from 7729 unique speakers. TalkVid is curated through a principled, multi-stage automated pipeline that rigorously filters for motion stability, aesthetic quality, and facial detail, and is validated against human judgments to ensure its reliability. Furthermore, we construct and release TalkVid-Bench, a stratified evaluation set of 500 clips meticulously balanced across key demographic and linguistic axes. Our experiments demonstrate that a model trained on TalkVid outperforms counterparts trained on previous datasets, exhibiting superior cross-dataset generalization. Crucially, our analysis on TalkVid-Bench reveals performance disparities across subgroups that are obscured by traditional aggregate metrics, underscoring its necessity for future research. Code and data can be found in https://github.com/FreedomIntelligence/TalkVid

cs.CV↗

arXiv2025

Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL

Recent advances in large language models (LLMs) and multi-agent systems have demonstrated remarkable capabilities in complex problem-solving tasks such as deep research, vibe coding, and mathematical reasoning. However, most existing multi-agent systems are built upon manual prompt/workflow engineering with sophisticated agent frameworks, making them computationally inefficient, less capable, and can not benefit from data-centric learning. In this work, we introduce Chain-of-Agents (CoA), a novel paradigm of LLM reasoning that enables native end-to-end complex problem-solving in the same way as a multi-agent system (i.e., multi-turn problem solving with multiple tools and multiple agents) within one model. In chain-of-agents problem-solving, the model dynamically activates different tool agents and role-playing agents to simulate multi-agent collaboration in an end-to-end fashion. To elicit end-to-end chain-of-agents problem-solving abilities in LLMs, we introduce a multi-agent distillation framework to distill state-of-the-art multi-agent systems into chain-of-agents trajectories for agentic supervised fine-tuning. We then use agentic reinforcement learning on verifiable agentic tasks to further improve the models' capabilities on chain-of-agents problem solving. We call the resulting models Agent Foundation Models (AFMs). Our empirical studies demonstrate that AFM establishes new state-of-the-art performance across diverse benchmarks in both web agent and code agent settings. We make the entire research, including the model weights, code for training and evaluation, and the training data, fully open-sourced, which offers a solid starting point for future research on agent models and agentic RL.

cs.AI↗

arXiv2025

MiCoTA: Bridging the Learnability Gap with Intermediate CoT and Teacher Assistants

Large language models (LLMs) excel at reasoning tasks requiring long thought sequences for planning, reflection, and refinement. However, their substantial model size and high computational demands are impractical for widespread deployment. Yet, small language models (SLMs) often struggle to learn long-form CoT reasoning due to their limited capacity, a phenomenon we refer to as the "SLMs Learnability Gap". To address this, we introduce \textbf{Mi}d-\textbf{Co}T \textbf{T}eacher \textbf{A}ssistant Distillation (MiCoTAl), a framework for improving long CoT distillation for SLMs. MiCoTA employs intermediate-sized models as teacher assistants and utilizes intermediate-length CoT sequences to bridge both the capacity and reasoning length gaps. Our experiments on downstream tasks demonstrate that although SLMs distilled from large teachers can perform poorly, by applying MiCoTA, they achieve significant improvements in reasoning performance. Specifically, Qwen2.5-7B-Instruct and Qwen2.5-3B-Instruct achieve an improvement of 3.47 and 3.93 respectively on average score on AIME2024, AMC, Olympiad, MATH-500 and GSM8K benchmarks. To better understand the mechanism behind MiCoTA, we perform a quantitative experiment demonstrating that our method produces data more closely aligned with base SLM distributions. Our insights pave the way for future research into long-CoT data distillation for SLMs.

cs.CL↗

arXiv2025

PersonaFeedback: A Large-scale Human-annotated Benchmark For Personalization

With the rapid improvement in the general capabilities of LLMs, LLM personalization, i.e., how to build LLM systems that can generate personalized responses or services that are tailored to distinct user personas, has become an increasingly important research and engineering problem. However, unlike many new challenging benchmarks being released for evaluating the general/reasoning capabilities, the lack of high-quality benchmarks for evaluating LLM personalization greatly hinders progress in this field. To address this, we introduce PersonaFeedback, a new benchmark that directly evaluates LLMs' ability to provide personalized responses given pre-defined user personas and queries. Unlike existing benchmarks that require models to infer implicit user personas from historical interactions, PersonaFeedback decouples persona inference from personalization, focusing on evaluating the model's ability to generate responses tailored to explicit personas. PersonaFeedback consists of 8298 human-annotated test cases, which are categorized into easy, medium, and hard tiers based on the contextual complexity of the user personas and the difficulty in distinguishing subtle differences between two personalized responses. We conduct comprehensive evaluations across a wide range of models. The empirical results reveal that even state-of-the-art LLMs that can solve complex real-world reasoning tasks could fall short on the hard tier of PersonaFeedback where even human evaluators may find the distinctions challenging. Furthermore, we conduct an in-depth analysis of failure modes across various types of systems, demonstrating that the current retrieval-augmented framework should not be seen as a de facto solution for personalization tasks. All benchmark data, annotation protocols, and the evaluation pipeline will be publicly available to facilitate future research on LLM personalization.

cs.CL↗

arXiv2025

A Reduction-based Algorithm for the Clique Interdiction Problem

The Clique Interdiction Problem (CIP) aims to minimize the size of the largest clique in a given graph by removing a given number of vertices. The CIP models a special Stackelberg game and has important applications in fields such as pandemic control and terrorist identification. However, the CIP is a bilevel graph optimization problem, making it very challenging to solve. Recently, data reduction techniques have been successfully applied in many (single-level) graph optimization problems like the vertex cover problem. Motivated by this, we investigate a set of novel reduction rules and design a reduction-based algorithm, RECIP, for practically solving the CIP. RECIP enjoys an effective preprocessing procedure that systematically reduces the input graph, making the problem much easier to solve. Extensive experiments on 124 large real-world networks demonstrate the superior performance of RECIP and validate the effectiveness of the proposed reduction rules.

cs.DS↗

arXiv2025

Is Your LLM Outdated? A Deep Look at Temporal Generalization

The rapid advancement of Large Language Models (LLMs) has led to the development of benchmarks that consider temporal dynamics, however, there remains a gap in understanding how well these models can generalize across temporal contexts due to the inherent dynamic nature of language and information. This paper introduces the concept of temporal generalization in LLMs, including bias in past and future generalizations. Then we introduce FreshBench, a new evaluation framework that employs fresh text and event prediction for assessing LLMs' temporal adaptability, ensuring the evaluation process free from data leakage and subjective bias. The experiment shows significant temporal biases and a decline in performance over time. Our findings reveal that powerful models, while initially superior, tend to decline more rapidly in future generalization. Additionally, powerful open-source models demonstrate better long-term adaptability compared to their closed-source counterparts. Our code is available at https://github.com/FreedomIntelligence/FreshBench.

cs.CL↗

arXiv2025

Cosmic Himalayas: The Highest Quasar Density Peak Identified in a 10,000 deg$^2$ Sky with Spatial Discrepancies between Galaxies, Quasars, and IGM HI

We report the identification of a quasar overdensity in the BOSSJ0210 field, dubbed Cosmic Himalayas, consisting of 11 quasars at $z=2.16-2.20$, the densest overdensity of quasars ($17σ$) in the $\sim$10,000 deg$^2$ of the Sloan Digital Sky Survey. We present the spatial distributions of galaxies and quasars and an HI absorption map of the intergalactic medium (IGM). On the map of 465 galaxies selected from the MAMMOTH-Subaru survey, we find two galaxy density peaks that do not fall on the quasar overdensity but instead exist at the northwest and southeast sides, approximately 25 $h^{-1}$ comoving-Mpc apart from the quasar overdensity. With a spatial resolution of 15 $h^{-1}$ comoving Mpc in projection, we produce a three-dimensional HI tomography map by the IGM Ly$α$ forest in the spectra of 23 SDSS/eBOSS quasars behind the quasar overdensity. Surprisingly, the quasar overdensity coincides with neither an absorption peak nor a transmission peak of IGM HI but lies near the border separating opaque and transparent volumes, with the more luminous quasars located in an environment with lesser IGM HI. Hence remarkably, the overdensity region traced by the 11 quasars, albeit all in coherently active states, has no clear coincidence with peaks of galaxies or HI absorption densities. Current physical scenarios with mixtures of HI overdensities and quasar photoionization cannot fully interpret the emergence of Cosmic Himalayas, suggesting this peculiar structure is an excellent laboratory to unveil the interplay between galaxies, quasars, and the IGM.

astro-ph.GA↗

arXiv2025

The Physical Origin of Extreme Emission Line Galaxies at High redshifts: Strong {\sc [Oiii]} Emission Lines Produced by Obscured AGNs

We present deep Subaru/FOCAS spectra for two extreme emission line galaxies (EELGs) at $z\sim 1$ with strong {\sc[Oiii]}$λ$5007 emission lines, exhibiting equivalent widths (EWs) of $2905^{+946}_{-578}$ Å and $2000^{+188}_{-159}$ Å, comparable to those of EELGs at high redshifts that are now routinely identified with JWST spectroscopy. Adding a similarly large {\sc [Oiii]} EW ($2508^{+1487}_{-689}$ Å) EELG found at $z\sim 2$ in the JWST CEERS survey to our sample, we explore for the physical origins of the large {\sc [Oiii]} EWs of these three galaxies with the Subaru spectra and various public data including JWST/NIRSpec, NIRCam, and MIRI data. While there are no clear signatures of AGN identified by the optical line diagnostics, we find that two out of two galaxies covered by the MIRI data show strong near-infrared excess in the spectral energy distributions (SEDs) indicating obscured AGN. Because none of the three galaxies show clear broad H$β$ lines, the upper limits on the flux ratios of broad-H$β$ to {\sc [Oiii]} lines are small, $\lesssim 0.15$ that are comparable with Seyfert $1.8-2.0$ galaxies. We conduct \texttt{Cloudy} modeling with the stellar and AGN incident spectra, allowing a wide range of parameters including metallicities and ionization parameters. We find that the large {\sc [Oiii]} EWs are not self-consistently reproduced by the spectra of stars or unobscured AGN, but obscured AGN that efficiently produces O$^{++}$ ionizing photons with weak nuclear and stellar continua that are consistent with the SED shapes.

astro-ph.GA↗

arXiv2024

MLLM-Bench: Evaluating Multimodal LLMs with Per-sample Criteria

Multimodal large language models (MLLMs) have broadened the scope of AI applications. Existing automatic evaluation methodologies for MLLMs are mainly limited in evaluating queries without considering user experiences, inadequately addressing the nuances of creative and associative multimodal tasks. However, the open-ended and subjective nature of such tasks poses a significant challenge to the evaluation methodology, where it is difficult to define the ground-truth answers for them. To this end, in our paper, we propose a new evaluation paradigm for MLLMs, which is evaluating MLLMs with per-sample criteria using potent MLLM as the judge. To validate the feasibility and effectiveness of this paradigm, we design a benchmark, dubbed MLLM-Bench, by curating the evaluation samples across six comprehensive cognitive levels. We benchmark 21 popular MLLMs in a pairwise-comparison fashion, showing diverse performance across models. Moreover, the validity of our benchmark manifests itself in reaching 88.02% agreement with human evaluation. We contend that the proposed paradigm explores the potential of MLLMs as effective evaluation tools with the help of per-sample criteria. See online leaderboard at \url{https://mllm-bench.llmzoo.com}.

cs.CL↗

arXiv2022

Improvement of cosmological constraints with the cross correlation between line-of-sight optical galaxy and FRB dispersion measure

Fast Radio Bursts (hereafter FRBs) can be used in cosmology by studying the Dispersion Measure (hereafter DM) as a function of redshift. The large scale structure of matter distribution is regarded as a major error budget for such application. Using optical galaxy and dispersion measure mocks built from N-body simulations, we have shown that the galaxy number density can be used as a tracer for large scale electron density and help improve the measurement of DM as a function of redshift. We have shown that, using the line-of-sight galaxy number counts within 1' around the given localized FRB source can help improve the cosmological parameter constraints by more than 20%.

astro-ph.CO↗

arXiv2019

XCMRC: Evaluating Cross-lingual Machine Reading Comprehension

We present XCMRC, the first public cross-lingual language understanding (XLU) benchmark which aims to test machines on their cross-lingual reading comprehension ability. To be specific, XCMRC is a Cross-lingual Cloze-style Machine Reading Comprehension task which requires the reader to fill in a missing word (we additionally provide ten noun candidates) in a sentence written in target language (English / Chinese) by reading a given passage written in source language (Chinese / English). Chinese and English are rich-resource language pairs, in order to study low-resource cross-lingual machine reading comprehension (XMRC), besides defining the common XCMRC task which has no restrictions on use of external language resources, we also define the pseudo low-resource XCMRC task by limiting the language resources to be used. In addition, we provide two baselines for common XCMRC task and two for pseudo XCMRC task respectively. We also provide an upper bound baseline for both tasks. We found that for common XCMRC task, translation-based method and multilingual sentence encoder-based method can obtain reasonable performance but still have much room for improvement. As for pseudo low-resource XCMRC task, due to strict restrictions on the use of language resources, our two approaches are far below the upper bound so there are many challenges ahead.

cs.CL↗