arXiv subjects

Songtao Huang

Publications and source records attributed to Songtao Huang.

18 recordsLinked to original sources

arXiv2026

Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent

We introduce Agents-A1, a 35B Mixture-of-Experts Agentic Model that reaches trillion-parameter-level performance by scaling the agent horizon. We investigate agent-horizon scaling from two perspectives: scaling long-horizon trajectories and scaling heterogeneous agent abilities. To support this goal, we build a long-horizon knowledge-action infrastructure that connects external knowledge, actions, observations, and verifier outcomes, producing agentic trajectories with an average length of 45K tokens. Based on this, we train Agents-A1 with a three-stage recipe. First, we perform full-domain supervised fine-tuning to align the base model with broad agentic behaviors. Second, we train domain-level teacher models to capture specialized expertise in each domain. Third, we propose a multi-teacher domain-routed on-policy distillation with salient vocabulary alignment to improve knowledge transfer efficiency across different domains, unifying six heterogeneous domains into one deployable student model. Agents-A1 achieves strong and broad performance for long-horizon agent benchmarks. Compared with 1T-parameter model such as Kimi-K2.6 and DeepSeek-V4-pro, Agents-A1 achieves leading results on SEAL-0 (56.4), IFBench (80.6), HiPhO (46.4), FrontierScience-Olympiad (79.0), and MolBench-Bind (56.8), and remains highly competitive on SciCode (44.3), HLE (47.6) and BrowseComp (75.5). We hope this work provides the community with a practical path for scaling the horizon using a 35B agent that can reach or match the performance of 1T models on long-horizon tasks.

cs.CL↗

arXiv2026

The Kubo-Thermalization Correspondence

Quantum thermalization describes how interacting quantum systems relax toward thermal equilibrium, a central problem in modern physics. Yet most experimental information on many-body systems comes from short-time transition spectroscopy, typically interpreted within Kubo's linear-response framework. These perspectives - long-time equilibration versus short-time response - seem fundamentally disconnected. Here we establish an exact link between them: the Kubo-Thermalization correspondence, which connects long-time thermalized magnetization under weak driving to short-time linear-response spectra for a spin coupled to a thermal bath. The correspondence holds even when the steady state differs substantially from the initial state and when each regime is individually difficult to describe theoretically. We experimentally confirm the correspondence using effective spin-1/2 impurities realized with ultracold fermions in two internal states coupled to a Fermi sea. Our results provide a rare exact statement about quantum thermalization and offer a novel route to infer thermalization dynamics from equilibrium response measurements in strongly interacting quantum systems, independent of microscopic details of the system-bath coupling.

cond-mat.quant-gas↗

arXiv2026

The Landscape of Agentic Reinforcement Learning for LLMs: A Survey

The emergence of agentic reinforcement learning (Agentic RL) marks a paradigm shift from conventional reinforcement learning applied to large language models (LLM RL), reframing LLMs from passive sequence generators into autonomous, decision-making agents embedded in complex, dynamic worlds. This survey formalizes this conceptual shift by contrasting the degenerate single-step Markov Decision Processes (MDPs) of LLM-RL with the temporally extended, partially observable Markov decision processes (POMDPs) that define Agentic RL. Building on this foundation, we propose a comprehensive twofold taxonomy: one organized around core agentic capabilities, including planning, tool use, memory, reasoning, self-improvement, and perception, and the other around their applications across diverse task domains. Central to our thesis is that reinforcement learning serves as the critical mechanism for transforming these capabilities from static, heuristic modules into adaptive, robust agentic behavior. To support and accelerate future research, we consolidate the landscape of open-source environments, benchmarks, and frameworks into a practical compendium. By synthesizing over five hundred recent works, this survey charts the contours of this rapidly evolving field and highlights the opportunities and challenges that will shape the development of scalable, general-purpose AI agents.

cs.AI↗

arXiv2026

Select-then-Solve: Paradigm Routing as Inference-Time Optimization for LLM Agents

When an LLM-based agent improves on a task, is the gain from the model itself or from the reasoning paradigm wrapped around it? We study this question by comparing six inference-time paradigms, namely Direct, CoT, ReAct, Plan-Execute, Reflection, and ReCode, across four frontier LLMs and ten benchmarks, yielding roughly 18,000 runs. We find that reasoning structure helps dramatically on some tasks but hurts on others: ReAct improves over Direct by 44pp on GAIA, while CoT degrades performance by 15pp on HumanEval. No single paradigm dominates, and oracle per-task selection beats the best fixed paradigm by 17.1pp on average. Motivated by this complementarity, we propose a select-then-solve approach: before answering each task, a lightweight embedding-based router selects the most suitable paradigm. Across four models, the router improves average accuracy from 47.6% to 53.1%, outperforming the best fixed paradigm at 50.3% by 2.8pp and recovering up to 37% of the oracle gap. In contrast, zero-shot self-routing only works for GPT-5 at 67.1% and fails for weaker models, all trailing the learned router. Our results argue that reasoning paradigm selection should be a per-task decision made by a learned router, not a fixed architectural choice.

cs.CL↗

arXiv2026

CoEnv: Driving Embodied Multi-Agent Collaboration via Compositional Environment

Multi-agent embodied systems hold promise for complex collaborative manipulation, yet face critical challenges in spatial coordination, temporal reasoning, and shared workspace awareness. Inspired by human collaboration where cognitive planning occurs separately from physical execution, we introduce the concept of compositional environment -- a synergistic integration of real-world and simulation components that enables multiple robotic agents to perceive intentions and operate within a unified decision-making space. Building on this concept, we present CoEnv, a framework that leverages simulation for safe strategy exploration while ensuring reliable real-world deployment. CoEnv operates through three stages: real-to-sim scene reconstruction that digitizes physical workspaces, VLM-driven action synthesis supporting both real-time planning with high-level interfaces and iterative planning with code-based trajectory generation, and validated sim-to-real transfer with collision detection for safe deployment. Extensive experiments on challenging multi-arm manipulation benchmarks demonstrate CoEnv's effectiveness in achieving high task success rates and execution efficiency, establishing a new paradigm for multi-agent embodied AI.

cs.RO↗

arXiv2026

Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse Supervision

Large language models (LLMs) have demonstrated exceptional reasoning capabilities, and co-evolving paradigms have shown promising results in domains such as code and math. However, in scientific reasoning tasks, these models remain fragile due to unreliable solution evaluation and limited diversity in verification strategies. In this work, we propose Sci-CoE, a two-stage scientific co-evolving framework that enables models to self-evolve as both solver and verifier through a transition from sparse supervision to unsupervised learning. In the first stage, the model uses a small set of annotated data to establish fundamental correctness judgment anchors for the Verifier. In the second stage, we introduce a geometric reward mechanism that jointly considers consensus, reliability, and diversity, driving large-scale self-iteration on unlabeled data. Experiments on several general scientific benchmarks demonstrate that Sci-CoE enhances complex reasoning capabilities and exhibits strong scalability, facilitating the construction of more robust and diverse evaluation systems. Codes are available at https://github.com/InternScience/Sci-CoE.

cs.AI↗

arXiv2026

InternAgent-1.5: A Unified Agentic Framework for Long-Horizon Autonomous Scientific Discovery

We introduce InternAgent-1.5, a unified system designed for end-to-end scientific discovery across computational and empirical domains. The system is built on a structured architecture composed of three coordinated subsystems for generation, verification, and evolution. These subsystems are supported by foundational capabilities for deep research, solution optimization, and long horizon memory. The architecture allows InternAgent-1.5 to operate continuously across extended discovery cycles while maintaining coherent and improving behavior. It also enables the system to coordinate computational modeling and laboratory experimentation within a single unified system. We evaluate InternAgent-1.5 on scientific reasoning benchmarks such as GAIA, HLE, GPQA, and FrontierScience, and the system achieves leading performance that demonstrates strong foundational capabilities. Beyond these benchmarks, we further assess two categories of discovery tasks. In algorithm discovery tasks, InternAgent-1.5 autonomously designs competitive methods for core machine learning problems. In empirical discovery tasks, it executes complete computational or wet lab experiments and produces scientific findings in earth, life, biological, and physical domains. Overall, these results show that InternAgent-1.5 provides a general and scalable framework for autonomous scientific discovery.

cs.AI↗

arXiv2025

Dispersion of first sound in a weakly interacting ultracold Fermi liquid

At low temperature, a normal gas of unpaired spin-1/2 fermions is one of the cleanest realizations of a Fermi liquid. It is described by Landau's theory, where no phenomenological parameters are needed as the quasiparticle interaction function can be computed perturbatively in powers of the scattering length $a$, the sole parameter of the short-range interparticle interactions. Obtaining an accurate solution of the transport equation nevertheless requires a careful treatment of the collision kernel, {as the uncontrolled error made by the relaxation time approximations increases when the temperature $T$ drops below the Fermi temperature}. Here, we study sound waves in the hydrodynamic regime up to second order in the Chapman-Enskog's expansion. We find that the frequency $ω_q$ of the sound wave is shifted above its linear departure as $ω_q=c_1 q(1+αq^2τ^2)$ where $c_1$ and $q$ are the speed and wavenumber of the sound wave and the typical collision time $τ$ scales as $1/a^2T^2$. Besides the shear viscosity, the coefficient $α$ is described by a single second-order collision time which we compute exactly from an analytical solution of the transport equation, resulting in a positive dispersion $α>0$. Our results suggest that ultracold atomic Fermi gases are an ideal experimental system for quantitative tests of second-order hydrodynamics.

cond-mat.quant-gas↗

arXiv2025

InternAgent: When Agent Becomes the Scientist -- Building Closed-Loop System from Hypothesis to Verification

Artificial Intelligence (AI) is accelerating the transformation of scientific research paradigms, not only enhancing research efficiency but also driving innovation. We introduce InternAgent, a unified closed-loop multi-agent framework to conduct Autonomous Scientific Research (ASR) across various scientific research fields, enabling researchers to tackle complicated problems in these fields with unprecedented speed and precision. InternAgent highlights three key advantages: 1) Scalability: InternAgent has demonstrated its versatility across 12 scientific research tasks, capable of generating innovative ideas to enhance the performance of baseline code. 2) Interactivity: InternAgent provides an interface for human expert feedback and multi-agent interaction in automated end-to-end processes, allowing for the seamless integration of domain expert knowledge. 3) Efficiency: InternAgent has achieved promising performance gains in several scientific fields with significantly less time cost compared to human efforts. For instance, in reaction yield prediction, it increased from 27.6% to 35.4% in just 12 hours; in enhancer activity prediction, accuracy rose from 0.65 to 0.79 with only 4 hours of processing; and in 2D semantic segmentation, precision advanced from 78.8% to 81.0% in a mere 30 hours.

cs.AI↗

arXiv2025

Emergence of Fermi's Golden Rule in the Probing of a Quantum Many-Body System

Fermi's Golden Rule (FGR) is one of the most impactful formulas in quantum mechanics, providing a link between easy-to-measure observables - such as transition rates - and fundamental microscopic properties - such as density of states or spectral functions. Its validity relies on three key assumptions: the existence of a continuum, an appropriate time window, and a weak coupling. Understanding the regime of validity of FGR is critical for the proper interpretation of most spectroscopic experiments. While the assumptions underlying FGR are straightforward to analyze in simple models, their applicability is significantly more complex in quantum many-body systems. Here, we observe the emergence and breakdown of FGR, using a strongly interacting homogeneous spin-$1/2$ Fermi gas coupled to a radio-frequency (rf) field. Measuring the transition probability into an outcoupled internal state, we map the system's dynamical response diagram versus the rf-pulse duration $t$ and Rabi frequency $Ω_0$. For weak drives, we identify three regimes: an early-time regime where the transition probability takes off as $t^2$, an intermediate-time FGR regime, and a long-time non-perturbative regime. Beyond a threshold Rabi frequency, Rabi oscillations appear. Our results provide a blueprint for the applicability of linear response theory to the spectroscopy of quantum many-body systems.

cond-mat.quant-gas↗

arXiv2025

TimeKAN: KAN-based Frequency Decomposition Learning Architecture for Long-term Time Series Forecasting

Real-world time series often have multiple frequency components that are intertwined with each other, making accurate time series forecasting challenging. Decomposing the mixed frequency components into multiple single frequency components is a natural choice. However, the information density of patterns varies across different frequencies, and employing a uniform modeling approach for different frequency components can lead to inaccurate characterization. To address this challenges, inspired by the flexibility of the recent Kolmogorov-Arnold Network (KAN), we propose a KAN-based Frequency Decomposition Learning architecture (TimeKAN) to address the complex forecasting challenges caused by multiple frequency mixtures. Specifically, TimeKAN mainly consists of three components: Cascaded Frequency Decomposition (CFD) blocks, Multi-order KAN Representation Learning (M-KAN) blocks and Frequency Mixing blocks. CFD blocks adopt a bottom-up cascading approach to obtain series representations for each frequency band. Benefiting from the high flexibility of KAN, we design a novel M-KAN block to learn and represent specific temporal patterns within each frequency band. Finally, Frequency Mixing blocks is used to recombine the frequency bands into the original format. Extensive experimental results across multiple real-world time series datasets demonstrate that TimeKAN achieves state-of-the-art performance as an extremely lightweight architecture. Code is available at https://github.com/huangst21/TimeKAN.

cs.LG↗

arXiv2025

Applying Ensemble Models based on Graph Neural Network and Reinforcement Learning for Wind Power Forecasting

Accurately predicting the wind power output of a wind farm across various time scales utilizing Wind Power Forecasting (WPF) is a critical issue in wind power trading and utilization. The WPF problem remains unresolved due to numerous influencing variables, such as wind speed, temperature, latitude, and longitude. Furthermore, achieving high prediction accuracy is crucial for maintaining electric grid stability and ensuring supply security. In this paper, we model all wind turbines within a wind farm as graph nodes in a graph built by their geographical locations. Accordingly, we propose an ensemble model based on graph neural networks and reinforcement learning (EMGRL) for WPF. Our approach includes: (1) applying graph neural networks to capture the time-series data from neighboring wind farms relevant to the target wind farm; (2) establishing a general state embedding that integrates the target wind farm's data with the historical performance of base models on the target wind farm; (3) ensembling and leveraging the advantages of all base models through an actor-critic reinforcement learning framework for WPF.

cs.LG↗

arXiv2024

Emergence of Sound in a Tunable Fermi Fluid

Landau's Fermi-liquid (FL) theory has been successful at the phenomenological description of the normal phase of many different Fermi systems. Using a dilute atomic Fermi fluid with tunable interactions, we investigate the microscopic basis of Landau's theory with a system describable from first principles. We study transport properties of an interacting Fermi gas by measuring its density response to a periodic external perturbation. In an ideal Fermi gas, we measure for the first time the celebrated Lindhard function. As the system is brought from the collisionless to the hydrodynamic regime, we observe the emergence of sound, and find that the experimental observations are quantitatively understood with a first-principle transport equation for the FL. When the system is more strongly interacting, we find deviations from such predictions. Finally, we observe the shape of the quasiparticle excitations directly from momentum-space tomography and see how it evolves from the collisionless to the collisional regime. Our study establishes this system as a clean platform for studying Landau's theory of the FL and paves the way for extending the theory to more exotic conditions, such as nonlinear dynamics and FLs with strong correlations in versatile settings.

cond-mat.quant-gas↗

arXiv2024

DST-GTN: Dynamic Spatio-Temporal Graph Transformer Network for Traffic Forecasting

Accurate traffic forecasting is essential for effective urban planning and congestion management. Deep learning (DL) approaches have gained colossal success in traffic forecasting but still face challenges in capturing the intricacies of traffic dynamics. In this paper, we identify and address this challenges by emphasizing that spatial features are inherently dynamic and change over time. A novel in-depth feature representation, called Dynamic Spatio-Temporal (Dyn-ST) features, is introduced, which encapsulates spatial characteristics across varying times. Moreover, a Dynamic Spatio-Temporal Graph Transformer Network (DST-GTN) is proposed by capturing Dyn-ST features and other dynamic adjacency relations between intersections. The DST-GTN can model dynamic ST relationships between nodes accurately and refine the representation of global and local ST characteristics by adopting adaptive weights in low-pass and all-pass filters, enabling the extraction of Dyn-ST features from traffic time-series data. Through numerical experiments on public datasets, the DST-GTN achieves state-of-the-art performance for a range of traffic forecasting tasks and demonstrates enhanced stability.

cs.AI↗

arXiv2023

The strongly driven Fermi polaron

Quasiparticles are emergent excitations of matter that underlie much of our understanding of quantum many-body systems. Therefore, the prospect of manipulating their properties with external fields -- or even destroying them -- has both fundamental and practical implications. However, in solid-state materials it is often challenging to understand how quasiparticles are modified by external fields owing to their complex interplay with other collective excitations, such as phonons. Here, we take advantage of the clean setting of homogeneous quantum gases and fast radio-frequency control to manipulate Fermi polarons -- quasiparticles formed by impurities interacting with a non-interacting Fermi gas -- from weak to ultrastrong drives. Exploiting two internal states of the impurity species, we develop a steady-state spectroscopy, from which we extract the energy of the driven polaron. We measure the decay rate and the quasiparticle residue of the driven polaron from the Rabi oscillations between the two internal states. At large drive strengths, the so-extracted quasiparticle residue exceeds unity, raising intriguing questions on the relationship between the Rabi oscillations and the impurity's spectral functions. Our experiment establishes the driven Fermi polaron as a promising platform for studying controllable quasiparticles in strongly driven quantum matter.

cond-mat.quant-gas↗

arXiv2023

Observation of the Fermionic Joule-Thomson Effect

We report the observation of the quantum Joule-Thomson (JT) effect in ideal and unitary Fermi gases. We study the temperature dynamics of these systems while they undergo an energy-per-particle conserving rarefaction. For scale-invariant systems, whose equations of state satisfy the relation $U\propto PV$, this rarefaction conserves the specific enthalpy, which makes it thermodynamically equivalent to a JT throttling process. We observe JT heating in an ideal Fermi gas, stronger at higher quantum degeneracy, a result of the repulsive quantum-statistical `force' arising from Pauli blocking. In a unitary Fermi gas, we observe that the JT heating is marginal in the temperature range $0.2 \lesssim T/T_{\mathrm{F}} \lesssim 0.8 $ as the repulsive quantum-statistical effect is lessened by the attractive interparticle interaction.

cond-mat.quant-gas↗

arXiv2023

Observation of Anomalous Decay of a Polarized Three-Component Fermi Gas

Systems of fermions with multiple internal states, such as quarks in quantum chromodynamics and nucleons in nuclear matter, are at the heart of some of the most complex quantum many-body problems. The stability of such many-body multi-component systems is crucial to understanding, for instance, baryon formation and the structure of nuclei, but these fermionic problems are typically very challenging to tackle theoretically. Versatile experimental platforms on which to study analogous problems are thus sought after. Here, we report the creation of a uniform gas of three-component fermions. We characterize the decay of this system across a range of interaction strengths and observe nontrivial competition between two- and three-body loss processes. We observe anomalous decay of the polarized (i.e. spin-population imbalanced) gas, in which the loss rates of each component unexpectedly differ. We introduce a generalized three-body rate equation which captures the decay dynamics, but the underlying microscopic mechanism is unknown.

cond-mat.quant-gas↗

arXiv2019

Entirety of Quantum Uncertainty and Its Experimental Verification

As a foundation of modern physics, uncertainty relations describe an ultimate limit for the measurement uncertainty of incompatible observables. Traditionally, uncertain relations are formulated by mathematical bounds for a specific state. Here we present a method for geometrically characterizing uncertainty relations as an entire area of variances of the observables, ranging over all possible input states. We find that for the pair of position $x$ and momentum $p$ operators, Heisenberg's uncertainty principle points exactly to the area of the variances of $x$ and $p$. Moreover, for finite-dimensional systems, we prove that the corresponding area is necessarily semialgebraic; in other words, this set can be represented via finite polynomial equations and inequalities, or any finite union of such sets. In particular, we give the analytical characterization of the areas of variances of (a) a pair of one-qubit observables, (b) a pair of projective observables for arbitrary dimension, and give the first experimental observation of such areas in a photonic system.

quant-ph↗