Searcharxiv⌕ Search

arXiv subjects

Chengcheng Han

Publications and source records attributed to Chengcheng Han.

At least 19 recordsLinked to original sources

Cosmological Collider Signals from Fermion Bubble Loops with a Chemical Potential

We present the first systematic analytical calculation of the local and nonlocal cosmological collider (CC) signals generated by massive fermion bubble loops in the presence of an effective chemical potential during inflation. Working in the de Sitter approximation, we investigate the four-point correlator of inflaton perturbations in the single-squeezed limit. By combining CC cutting rules, late-time expansions, and partial Mellin-Barnes (MB) techniques, we derive analytical expressions for both local and nonlocal signals. In particular, we identify two distinct contributions to the local signal, arising from spectrum and integer poles and from ultraviolet poles of the loop-momentum integral. We further investigate the three-point limit, where the separately extracted local and nonlocal contributions develop apparent folded singularities. By reorganizing the MB representation, we demonstrate that these singularities cancel in the complete signal and obtain a convergent double-MB integral suitable for numerical evaluation. Independent calculations using complementary MB representations provide nontrivial consistency checks of our results. We find that the chemical potential can significantly enhance the oscillatory signal while increasing its logarithmic oscillation frequency. Our analytical results provide a foundation for further studies of chemical-potential-enhanced fermionic CC signals, including fermion triangle loops.

hep-ph↗

TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training

Training tool-use agents typically relies on outcome-based filtering: Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks. However, this paradigm ignores interaction dynamics: successful trajectories may lack error recovery or exhibit redundancy, while pass rates fail to distinguish structurally informative tasks from trivial ones. We propose \textbf{TopoCurate}, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology. By merging equivalent action-observation states, this projection transforms scattered linear trajectories into a structured manifold that explicitly captures how tool invocations and environmental responses drive the divergence between effective strategies and failure modes. Leveraging this representation, we introduce a dual-selection mechanism: for SFT, we prioritize trajectories demonstrating reflective recovery, semantic efficiency, and strategic diversity to mitigate covariate shift and mode collapse; for RL, we select tasks with high error branch ratios and strategic heterogeneity, maximizing gradient Signal-to-Noise Ratio to address vanishing signals in sparse-reward settings. Evaluations on BFCLv3 and Tau2 Bench show that TopoCurate achieves consistent gains of 4.2\% (SFT) and 6.9\% (RL) over state-of-the-art baselines. We will release the code and data soon for further investigations.

cs.LG↗

DORA: A Scalable Asynchronous Reinforcement Learning System for Language Model Training

Reinforcement learning (RL) has become a critical paradigm for LLM post-training, yet the rollout phase -- accounting for 50--80% of total step time -- is bottlenecked by skewed generation: long-tailed trajectories indispensable for model performance block the entire training pipeline. Asynchronous training offers a natural remedy by overlapping generation with training, but introduces a fundamental tension between efficiency and algorithmic correctness. We identify three constraints in asynchronous training to preserve convergence: intra-trajectory policy consistency, data integrity, and bounded staleness. Existing approaches fail to intrinsically address the long-tailed trajectory problem, which is further exacerbated by the imbalance characteristic of Mix-of-Experts models, or deviate from the standard RL training formulation, thereby hindering model convergence. Therefore, we propose DORA (Dynamic ORchestration for Asynchronous Rollout), which addresses this challenge through algorithm-system co-design. DORA introduces multi-version streaming rollout, a novel asynchronous paradigm that maintains multiple policy versions concurrently -- simultaneously achieving full bubble elimination without compromising algorithmic constraints. Experimental results demonstrate that our DORA system achieves substantial improvements in throughput -- up to 2--3 times higher than state-of-the-art systems on open-source benchmarks -- without compromising convergence. Furthermore, in large-scale industrial applications with tens of thousands of accelerators, DORA accelerates RL training by 2--4 times compared to synchronous training across various scenarios. The resultant open-source models, LongCat-Flash-Thinking, exhibit competitive performance on complex reasoning benchmarks, matching the capability of most advanced LLMs.

cs.LG↗

Right-Handed Neutrino Production by an Axion-like Inflaton: Implications for Leptogenesis

We study heavy right-handed neutrino production induced by a derivative coupling to an axion-like inflaton and its implications for non-thermal leptogenesis. We develop a unified treatment of Majorana-fermion production during inflation and preheating. During inflation, the rolling inflaton background generates a helicity-asymmetric spectrum that can be obtained analytically in the slow-roll regime, while during preheating the oscillating background drives repeated non-adiabatic production events with a helicity-dependent effective momentum and a natural momentum cutoff. We compute the resulting abundance in both the delayed-decay and prompt-decay limits, including the effects of Pauli blocking and successive production stages. We further clarify the relation between the fermion basis used to identify non-adiabatic production and the Hamiltonian-diagonal basis used to define instantaneous occupation numbers. Applying these results to non-thermal leptogenesis, we identify parameter regions consistent with the observed baryon asymmetry.

hep-ph↗

Cosmological Collider Signatures from Right-Handed Neutrino Loop

We study cosmological collider (CC) signatures generated by right-handed neutrino loops in the setup of inflation combined with the neutrino seesaw mechanism. We formulate the inflaton interaction with the right-handed neutrinos through a unique dimension-5 operator respecting shift symmetry, which induces an effective chemical potential in the slow-roll background, leading to helicity-dependent production of right-handed neutrinos and enhancing the CC signatures. The right-handed neutrino is described by a two-component Weyl spinor with a Majorana mass term. Using the Schwinger-Keldysh formalism, we derive a set of seed integrals for fermion propagators of the right-handed Majorana neutrino. With these, we compute the factorized nonlocal contributions to the three-point inflaton correlator generated by the triangle loop of right-handed neutrinos. We show that the chemical potential can substantially soften the heavy-mass Boltzmann suppression and amplify the oscillatory non-Gaussianity signatures associated with the dominant helicity mode. We systematically analyze the spinor structure, the factorization of fermion triangle loop, the loop-momentum integrals and the time seed integrals, which give a controlled computation of the nonlocal contributions to the CC signals of the right-handed neutrinos.

hep-ph↗

Line-of-sight magnetic-field propagation effects on axion-like particle constraints from GRB 221009A

High-energy photons from GRB 221009A provide a powerful opportunity to probe axion-like particles (ALPs) through photon-ALP oscillations in cosmic magnetic fields. We revisit the ALP constraints implied by the LHAASO observation of this burst, with particular emphasis on the magnetic-field environments encountered along the line of sight. We include the host-galaxy, intergalactic, and Milky-Way magnetic fields and assess their respective impacts on the photon survival probability and on the exclusion limits in the ALP mass-coupling plane. We show that the constraints are only mildly affected by the choice of host-galaxy and Galactic magnetic-field models, but can change significantly once the intergalactic magnetic field is varied. Its field strength, coherence scale, and stochastic properties can all leave visible imprints on the derived exclusion contours, and in some cases generate pronounced oscillatory features. This demonstrates that the intergalactic magnetic field constitutes the dominant astrophysical uncertainty in extracting ALP limits from GRB 221009A. Our analysis highlights the importance of realistic propagation modeling in future gamma-ray searches for ALPs.

hep-ph↗

Skill0.5: Joint Skill Internalization and Utilization for Out-of-Distribution Generalization in Agentic Reinforcement Learning

Equipping large language models with explicit skills has emerged as a promising paradigm for enabling autonomous agents to solve complex tasks. Agent skills can be inherently divided into general skills for broad cognitive transfer and task-specific skills for dynamic execution. However, existing skill-based reinforcement learning (RL) methods typically force a rigid choice between full externalization, which incurs prohibitive context overhead, and full internalization, which risks overfitting and knowledge conflicts. To address this dilemma, we propose Skill0.5, a novel agentic RL framework that explicitly differentiates skill treatments by combining general skill internalization with task-specific skill utilization. Driven by a dynamic, difficulty-aware router, Skill0.5 streams tasks into distinct mastery tiers to apply tailored optimization strategies: it internalizes general skills via privileged distillation to build a cognitive foundation for hard tasks, while using diagnostic probing on easy tasks to penalize shortcuts and enforce specific skill utilization. Experiments on ALFWorld and WebShop demonstrate that Skill0.5 outperforms both memory-based and skill-based RL baselines, yielding performance improvements across both in-distribution and out-of-distribution scenarios.

cs.CL↗

GUI-CIDER: Mid-training GUI Agents via Causal Internalization and Density-aware Exemplar Reselection

Despite the rapid progress of multimodal large language models in building Graphical User Interface (GUI) agents, their real-world task completion is fundamentally bottlenecked by a lack of world knowledge about GUI operations. Existing solutions typically rely on expensive multi-agent scaffolding or conventional post-training paradigms, such as Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL). However, post-training only allows agents to implicitly absorb world knowledge through action annotations or reward signals, leading to inefficient trajectory memorization rather than genuine comprehension. Therefore, an approach that enables explicit learning of this knowledge is imperative. To this end, we propose GUI-CIDER, a mid-training method that explicitly internalizes GUI world knowledge through Causal Internalization and Density-aware Exemplar Reselection. GUI-CIDER operates in three stages: (1) data synthesis, which distills static planning and dynamic causal knowledge from GUI trajectories into text; (2) exemplar reselection, which filters the corpus by rewarding causal structures and penalizing semantic redundancy; and (3) mid-training, where the refined data is used to embed the acquired knowledge. Extensive experiments on two GUI knowledge benchmarks and three task completion benchmarks demonstrate that GUI-CIDER consistently improves both the agent's understanding of GUI operations and its task success rates.The codes are available at https://github.com/Wuzheng02/GUI-CIDER.

cs.CL↗

SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization

Agent skills, structured packages of procedural knowledge and executable resources that agents dynamically load at inference time, have become a reliable mechanism for augmenting LLM agents. Yet inference-time skill augmentation is fundamentally limited: retrieval noise introduces irrelevant guidance, injected skill content imposes substantial token overhead, and the model never truly acquires the knowledge it merely follows. We ask whether skills can instead be internalized into model parameters, enabling zero-shot autonomous behavior without any runtime skill retrieval. We introduce SKILL0, an in-context reinforcement learning framework designed for skill internalization. SKILL0 introduces a training-time curriculum that begins with full skill context and progressively withdraws it. Skills are grouped offline by category and rendered with interaction history into a compact visual context, teaching he model tool invocation and multi-turn task completion. A Dynamic Curriculum then evaluates each skill file's on-policy helpfulness, retaining only those from which the current policy still benefits within a linearly decaying budget, until the agent operates in a fully zero-shot setting. Extensive agentic experiments demonstrate that SKILL0 achieves substantial improvements over the standard RL baseline (+9.7\% for ALFWorld, +6.6\% for Search-QA, and+10.1\% for WebShop), while maintaining a highly efficient context of fewer than 0.5k tokens per step. Our code is available at https://github.com/ZJU-REAL/SkillZero.

cs.LG↗

LongCat-Flash-Thinking-2601 Technical Report

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

cs.AI↗

Curvature Perturbations from Higgs Modulated Reheating

In this work we investigate curvature perturbations and non-Gaussianity arising from Higgs modulated reheating in the early Universe. We employ three different methods -- the period-averaging (PA) method, the exact method, and the non-perturbative $δN$ formalism -- to compute the power spectrum and bispectrum of curvature perturbations. Our results show that the non-perturbative $δN$ method provides a reliable estimate across a wide range of reheating time and Higgs field values, including regimes where the Higgs field oscillates significantly after inflation. We find that a smaller Higgs self-coupling ($λ$) leads to a larger curvature perturbation, with the non-Gaussianity predominantly taking a local shape. This highlights the importance of considering non-perturbative effects in calculating the curvature perturbation during Higgs modulated reheating, especially for smaller values of $λ$. Our findings offer valuable insights into the dynamics of reheating and the generation of primordial perturbations in the early Universe.

astro-ph.CO↗

Introducing LongCat-Flash-Thinking: A Technical Report

We present LongCat-Flash-Thinking, an efficient 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model. Its advanced capabilities are cultivated through a meticulously crafted training process, beginning with long Chain-of-Thought (CoT) data cold-start and culminating in large-scale Reinforcement Learning (RL). We first employ a well-designed cold-start training strategy, which significantly enhances the reasoning potential and equips the model with specialized skills in both formal and agentic reasoning. Then, a core innovation is our domain-parallel training scheme, which decouples optimization across distinct domains (e.g., STEM, Code, Agentic) and subsequently fuses the resulting expert models into a single, nearly Pareto-optimal model. This entire process is powered by our Dynamic ORchestration for Asynchronous rollout (DORA) system, a large-scale RL framework that delivers a greater than threefold training speedup over synchronous methods on tens of thousands of accelerators. As a result, LongCat-Flash-Thinking achieves state-of-the-art performance among open-source models on a suite of complex reasoning tasks. The model exhibits exceptional efficiency in agentic reasoning, reducing average token consumption by 64.5% (from 19, 653 to 6, 965) on AIME-25, without degrading task accuracy. We release LongCat-Flash-Thinking to promote further advances in reasoning systems and agentic AI research.

cs.AI↗

Cosmological Signatures of Neutrino Seesaw Mechanism

The tiny neutrino masses are most naturally explained by seesaw mechanism through singlet right-handed neutrinos, which can further explain the matter-antimatter asymmetry in the Universe. In this Letter, we propose a new approach to study cosmological signatures of neutrino seesaw through the interaction between inflaton and right-handed neutrinos that respects the shift symmetry. In our framework, after inflation the inflaton predominantly decays into right-handed neutrinos and its decay rate is modulated by the fluctuations of Higgs field that act as the source of curvature perturbations. This gives a new realization of Higgs modulated reheating, and it produces primordial non-Gaussian signatures that can be measured by the forthcoming large-scale structure surveys. We find that these surveys have the potential to probe a large portion of the neutrino seesaw parameter space, opening up a new window for testing the high scale seesaw mechanism.

hep-ph↗

Cosmological Non-Gaussianity from Neutrino Seesaw Mechanism

The neutrino mass generation via conventional seesaw mechanism is realized at high scales around $O(10^{14})$GeV with natural Yukawa couplings of $O(1)$, making the test of neutrino seesaw a great challenge. It is intriguing to note that the neutrino seesaw scale is typically around the upper range of the cosmological inflation scale. In this work, we propose a new framework incorporating inflation and neutrino seesaw in which the inflaton primarily decays into right-handed neutrinos after inflation. This decay process is governed by the inflaton interaction with the right-handed neutrinos that respects the shift symmetry. With the neutrino seesaw mechanism, we construct a new realization of the Higgs modulated reheating, in which the fluctuations of Higgs field can modulate the inflaton decays and contribute to the primordial curvature perturbation. We investigate the induced non-Gaussian signatures and demonstrate, for the first time, that such signatures provide an important means to directly probe the high scale of natural neutrino seesaw. We further analyze the interplay of the non-Gaussianity signatures with the low-energy neutrino experiments, and their interplay with the Higgs self-coupling measurements at the LHC and future colliders.

hep-ph↗

VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications

As LLM-based agents are increasingly deployed in real-life scenarios, existing benchmarks fail to capture their inherent complexity of handling extensive information, leveraging diverse resources, and managing dynamic user interactions. To address this gap, we introduce VitaBench, a challenging benchmark that evaluates agents on versatile interactive tasks grounded in real-world settings. Drawing from daily applications in food delivery, in-store consumption, and online travel services, VitaBench presents agents with the most complex life-serving simulation environment to date, comprising 66 tools. Through a framework that eliminates domain-specific policies, we enable flexible composition of these scenarios and tools, yielding 100 cross-scenario tasks (main results) and 300 single-scenario tasks. Each task is derived from multiple real user requests and requires agents to reason across temporal and spatial dimensions, utilize complex tool sets, proactively clarify ambiguous instructions, and track shifting user intent throughout multi-turn conversations. Moreover, we propose a rubric-based sliding window evaluator, enabling robust assessment of diverse solution pathways in complex environments and stochastic interactions. Our comprehensive evaluation reveals that even the most advanced models achieve only 30% success rate on cross-scenario tasks, and less than 50% success rate on others. Overall, we believe VitaBench will serve as a valuable resource for advancing the development of AI agents in practical real-world applications. The code, dataset, and leaderboard are available at https://vitabench.github.io/

cs.CL↗

New Physics Search at the CEPC: a General Perspective

The Circular Electron-Positron Collider (CEPC), a proposed next-generation Higgs factory, provides new opportunities to explore physics beyond the Standard Model (SM). With its clean electron-positron collision environment and the ability to collect large samples of Higgs, W, and Z bosons, the CEPC enables precision measurements and searches for new physics. This white paper outlines the CEPC's discovery potential, including studies of exotic decays of the Higgs, Z, and top quarks, dark matter and dark sector phenomena, long-lived particles, supersymmetry, and neutrino-related signatures. Advanced detector technologies and reconstruction techniques, such as one-to-one correspondence reconstruction and jet origin identification, significantly improve sensitivity to rare and weakly interacting processes. The CEPC is particularly well suited to probe the electroweak phase transition and test models of electroweak baryogenesis and dark sector interactions. In addition, global fit analyses highlight the CEPC's complementary role in constraining a wide range of new physics scenarios. These features position the CEPC as a powerful tool for exploring the next frontier in fundamental particle physics in the post-Higgs discovery era.

hep-ex↗

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

MUSE: MCTS-Driven Red Teaming Framework for Enhanced Multi-Turn Dialogue Safety in Large Language Models

As large language models~(LLMs) become widely adopted, ensuring their alignment with human values is crucial to prevent jailbreaks where adversaries manipulate models to produce harmful content. While most defenses target single-turn attacks, real-world usage often involves multi-turn dialogues, exposing models to attacks that exploit conversational context to bypass safety measures. We introduce MUSE, a comprehensive framework tackling multi-turn jailbreaks from both attack and defense angles. For attacks, we propose MUSE-A, a method that uses frame semantics and heuristic tree search to explore diverse semantic trajectories. For defense, we present MUSE-D, a fine-grained safety alignment approach that intervenes early in dialogues to reduce vulnerabilities. Extensive experiments on various models show that MUSE effectively identifies and mitigates multi-turn vulnerabilities. Code is available at \href{https://github.com/yansiyu02/MUSE}{https://github.com/yansiyu02/MUSE}.

cs.CL↗