arXiv subjects

Tian Liang

Publications and source records attributed to Tian Liang.

At least 19 recordsLinked to original sources

arXiv2026

MetaRAG: Belief-Action Aligned Policy Optimization for Agentic RAG

Agentic retrieval-augmented generation (RAG) requires language models to decide when to continue searching and when to answer. Existing RL-based methods rely on external supervision and overlook the agent's internal belief about whether the current evidence is sufficient. To address this problem, we reformulate the search decision quality as belief-action alignment and propose MetaRAG, a belief-action aligned policy optimization framework for agentic RAG. MetaRAG uses Verify-first Action Generation to elicit an explicit verification process before each actual action, and Internal Belief Probing to estimate the policy model's own answerability belief from the same question-history context. Based on these, MetaRAG derives a consistency reward that is further gated by answer correctness, avoiding reinforcement of internally consistent but incorrect trajectories. The belief probe is used only during training and introduces no inference-time overhead. Experiments on seven public QA benchmarks show that MetaRAG consistently improves the accuracy-efficiency trade-off over strong RL-based agentic RAG baselines, with gains that transfer to deep research settings, different optimizers, and multiple model backbones.

cs.AI↗

arXiv2026

CARL: Constraint-Aware Reinforcement Learning for Planning with LLMs

Despite their strong reasoning capabilities and extensive world knowledge, Large Language Models (LLMs) frequently generate plans that violate task constraints, undermining their reliability in real-world applications. This deficiency arises from a lack of systematic mechanisms to incorporate constraint information during the generation process. While existing approaches attempt to mitigate this by relying on external tools or task decomposition, they fail to enhance the model's intrinsic constraint awareness. To address this, we propose Constraint-Aware Reinforcement Learning (CARL), a novel RL framework designed to strengthen LLMs' intrinsic focus on constraints. CARL introduces a constraint-aware reward by comparing the model's output distributions under constrained and unconstrained inputs, encouraging constraint focus and penalizing neglect. Compatible with various RL frameworks and requiring no external solvers or top models, CARL enables scalable, end-to-end constraint-aware planning. Extensive experiments on BlocksWorld, TravelPlanner, and T-Eval demonstrate that CARL significantly outperforms standard Reinforcement Fine-Tuning (RFT) baselines and state-of-the-art reasoning models, exhibiting a markedly increased focus on constraints.

cs.AI↗

arXiv2026

STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training

Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task goal and interaction history at intermediate steps. Prior work has explored step-level supervision using Shannon-entropy-based uncertainty signals, which conflate inherent state complexity with agent confidence and therefore provide unreliable estimates of decision reliability. To address this issue, we propose normalized entropy, which measures confidence deviations relative to an agent's average behavior under a given state, thereby strengthening the association between low-quality actions and trajectory neglect. Building on this insight, we introduce Selective Trajectory-Aware Policy Optimization (STAPO), a hierarchical group-based RL framework. STAPO leverages normalized entropy to locate outlier steps associated with trajectory neglect and optimizes them via a joint mechanism of trajectory-aware reward and trajectory-independent penalty, enhancing trajectory awareness while preserving training stability. Extensive experiments on ALFWorld, WebShop, and Search-Augmented QA demonstrate that STAPO achieves state-of-the-art performance while substantially alleviating trajectory neglect, validating its effectiveness and robustness for agentic tasks.

cs.AI↗

arXiv2026

Hierarchical Sparse Attention Done Right: Toward Infinite Context Modeling

Scaling modern large language models (LLMs) to long contexts is limited by the quadratic computation cost, and poor length extrapolation of dense attention. Chunk-wise sparse attention offers a promising alternative, but all existing methods fall short of full attention because of their inaccurate chunk selection. We propose Hierarchical Landmark Sparse (HiLS) Attention, a chunk-wise sparse attention mechanism that learns chunk selection end-to-end under the language-modeling (LM) loss. HiLS factorizes attention hierarchically: each query performs attention independently with each retrieved chunk to extract chunk-specific information, and the resulting outputs are fused according to chunk retrieval scores. By incorporating retrieval scores into the forward attention computation, HiLS optimizes them directly with the LM loss, enabling end-to-end retrieval learning and native sparse training. Experimental results show that HiLS-Attention achieves performance comparable to, and in some cases better than, full attention at in-domain context lengths. Meanwhile, HiLS-Attention extrapolates more than $64\times$ the training context length with 90% retrieval accuracy, far beyond full attention. Moreover, existing full-attention models can be converted to HiLS-Attention with lightweight continued pretraining, preserving in-domain performance while acquiring ultra-long-context extrapolation. Together with its sparse KV access and computation, HiLS-Attention breaks the usual efficiency-performance trade-off, enabling long-context LLMs that are both more efficient and more effective on general long-context tasks than their full-attention counterparts.

cs.CL↗

arXiv2026

Confidence Calibration for Multimodal LLMs: An Empirical Study through Medical VQA

Multimodal Large Language Models (MLLMs) show great potential in medical tasks, but their elicited confidence often misaligns with actual accuracy, potentially leading to misdiagnosis or overlooking correct advice. This study presents the first comprehensive analysis of the relationship between accuracy and confidence in medical MLLMs. It proposes a novel method that combines Multi-Strategy Fusion-Based Interrogation (MS-FBI) with auxiliary expert LLM assessment, aiming to improve confidence calibration in Medical Visual Question Answering (VQA). Experiments demonstrate that our method reduces the Expected Calibration Error (ECE) by an average of 40\% across three Medical VQA datasets, significantly enhancing MLLMs' reliability. The findings highlight the importance of domain-specific calibration for MLLMs in healthcare, offering a more trustworthy solution for AI-assisted diagnosis.

cs.CV↗

arXiv2026

Topological Surface Charge Detection via Terahertz Time-domain Spectroscopy

The topological magnetoelectric effect (TME) in three-dimensional topological insulators manifests as a quantized surface charge accumulation proportional to an applied magnetic field. Here we demonstrate an optical method using terahertz time-domain spectroscopy (THz-TDS) to detect surface charge accumulation in a chromium-doped (Bi,Sb)$_2$Te$_3$ thin film under oblique incidence, achieving sub-milliradian Faraday rotation precision. Unlike transport probes that require ultralow longitudinal conductivity, this optical technique is robust against finite $\sigma_L$, degrading by less than $0.3\%$ even when $\sigma_L \sim \sigma_T$. We extract the charge accumulation $\eta/B_z$ from the measured Faraday rotation and show results at $45^\circ$ and $60^\circ$ coincide within experimental uncertainty. Extending this to axion insulators, we predict that the TME produces an imaginary Faraday rotation linear in frequency, whose slope directly reflects the single-surface charge density. With improved sample thickness and precision, this optical scheme provides a viable pathway toward direct verification of the TME and four-dimensional quantum Hall effect.

cond-mat.mes-hall↗

arXiv2026

FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention

Conventional LLMs keep the full KV cache loaded during decoding, causing a severe GPU memory bottleneck for ultra-long context serving. In this report, we propose \textbf{Lookahead Sparse Attention (LSA)}, a novel inference paradigm powered by a Neural Memory Indexer built upon the DeepSeek-V4 architecture. Rather than passively attending to all historical tokens, LSA proactively predicts future context demands and preserves only the query-critical KV chunks in the GPU memory. Crucially, we instantiate this architecture via a \textbf{backbone-free decoupled training} strategy. By formulating the indexer as a standard dual-encoder architecture, we train it independently using standard retrieval training frameworks without ever loading the massive backbone model into GPU memory. We demonstrate that this ``less is more'' paradigm significantly maximizes serving efficiency while acting as an effective attention denoiser in tasks that rely on long-term global memory. Across primary long-context evaluation suites (e.g., LongBench-v2, LongMemEval, and RULER), \texttt{FM-DS-V4} compresses the average physical KV cache footprint down to merely 13.5\% of the full-context baseline, while consistently preserving or slightly elevating downstream accuracy (+0.6\% absolute margin on average). At 1M context, per-decode-token compute drops to 0.30$\times$ of the baseline and GPU KV cache shrinks by 90\% (3.73$\to$0.37 GB), translating into \textbf{2.8$\times$ aggregate throughput and 2.7$\times$ concurrency gains} in PD-disaggregated serving on 8$\times$H20 GPUs.

cs.LG↗

arXiv2026

Towards Robotic Dexterous Hand Intelligence: A Survey

Robotic dexterous hands are central to contact-rich manipulation, with rapid progress driven by advances in hardware, sensing, control, simulation, and data generation. However, existing studies are often developed under different assumptions regarding hand embodiments, sensory configurations, task settings, training data, and evaluation protocols, making systematic comparison difficult and obscuring the developmental trajectory of the field. This survey provides a holistic review of dexterous hand research from four complementary aspects. First, we present a hardware-level analysis covering actuation, transmission, perception, and representative hand designs, highlighting the key trade-offs in force capability, compliance, bandwidth, integration, and system complexity. Furthermore, we review control and learning methods for dexterous manipulation from a methodological perspective, grouping representative works by major paradigms and tracing their evolution in chronological order. In addition, we consolidate datasets, modality design, and evaluation practices, which enables methodological progress to be interpreted together with the ways in which it is trained, benchmarked, and assessed. Finally, we discuss the major limitations of current dexterous hand research and summarize the corresponding future directions. By connecting hardware analysis, methodological development, data resources, and evaluation, this survey aims to provide a structured understanding of dexterous hand research and to clarify the most important open challenges for future study.

cs.RO↗

arXiv2026

From $P(y|x)$ to $P(y)$: Investigating Reinforcement Learning in Pre-train Space

While reinforcement learning with verifiable rewards (RLVR) significantly enhances LLM reasoning by optimizing the conditional distribution P(y|x), its potential is fundamentally bounded by the base model's existing output distribution. Optimizing the marginal distribution P(y) in the Pre-train Space addresses this bottleneck by encoding reasoning ability and preserving broad exploration capacity. Yet, conventional pre-training relies on static corpora for passive learning, leading to a distribution shift that hinders targeted reasoning enhancement. In this paper, we introduce PreRL (Pre-train Space RL), which applies reward-driven online updates directly to P(y). We theoretically and empirically validate the strong gradient alignment between log P(y) and log P(y|x), establishing PreRL as a viable surrogate for standard RL. Furthermore, we uncover a critical mechanism: Negative Sample Reinforcement (NSR) within PreRL serves as an exceptionally effective driver for reasoning. NSR-PreRL rapidly prunes incorrect reasoning spaces while stimulating endogenous reflective behaviors, increasing transition and reflection thoughts by 14.89x and 6.54x, respectively. Leveraging these insights, we propose Dual Space RL (DSRL), a Policy Reincarnation strategy that initializes models with NSR-PreRL to expand the reasoning horizon before transitioning to standard RL for fine-grained optimization. Extensive experiments demonstrate that DSRL consistently outperforms strong baselines, proving that pre-train space pruning effectively steers the policy toward a refined correct reasoning subspace.

cs.LG↗

arXiv2026

Optimized growth of large-size, high quality $\text{ZrTe}_5$ single crystals enabling clear quantum oscillations in electrical transport

Quantum oscillation with nontrivial Berry phase is one of the characteristics of topological materials. As a Dirac semimetal candidate, zirconium pentatelluride ($\text{ZrTe}_5$) stands out as an intriguing material for investigating topological phase transitions and Dirac fermion physics; however, the extreme sensitivity of its electronic properties to stoichiometric variations and crystalline defects has hindered consistent experimental observation. Here, we report an optimized Te-flux synthesis method designed to produce centimeter-scale, high-quality single crystals meanwhile minimizing extrinsic carrier contamination. Comprehensive morphology, structural and chemical characterizations, including scanning electron microscopy, Laue backscattering and Rietveld refinement, confirm a high-purity $Cmcm$ phase with excellent crystallinity. Furthermore, magnetotransport measurements reveal a remarkably low Shubnikov-de Haas oscillation onset field ($B_{int} \approx 0.38$ T) with an ultra-high mobility of $5.58\times10^5$cm$^2$V$^{-1}$s$^{-1}$ and access to the the quantum limit at $B \approx 1.3$ T, attesting to the superior crystalline quality and the efficacy of this growth optimization. These results demonstrate that growth control is crucial for stabilizing intrinsic electronic behavior in $\text{ZrTe}_5$, establishing a robust platform for exploring topological phase transitions and exotic quantum phenomena in topological semimetals.

cond-mat.mtrl-sci↗

arXiv2026

Observation of Superfluidity and Meissner Effect of Composite Bosons in GaAs Quantum Hall System

The quantum Hall effect (QHE) is theoretically understood as a superfluid condensate of composite bosons (CBs) -- bound states of electrons and magnetic flux quanta. While dissipationless transport is consistent with this picture, other signatures of superfluidity, such as the Meissner effect, remain elusive. Here, we present direct experimental evidence for CB superfluidity by probing the system's response to a controlled, time-varying magnetic field in Corbino disk geometries. We simultaneously observe the quantized Laughlin charge pumping and a new, quantized charge accumulation phenomenon, governed by the relation $\Delta Q_{\rm a}/e = \nu\,(\Delta \Phi/\Phi_0)$. This relation signifies that the system actively maintains the fixed electron-to-flux ratio that defines the CBs, neutralizing excess flux by drawing in a precise number of electrons. Crucially, devices with multiple concentric top gates reveal that this charge accumulation is uniformly distributed across the bulk of the QHE fluid, demonstrating that it is a collective, bulk property rather than an edge effect -- a key signature of a superfluid condensate. Furthermore, the presence of a top gate determines the screening mechanism: in a "grand canonical" setting with a gate, low Coulomb energy favors a charge-mediated screening (generalized Meissner effect); without a gate, the system enters a "canonical" regime, exhibiting fixed electron density like type-II superconductors. These observations confirm the CB superfluid nature of the QHE ground state and establish a versatile platform for studying macroscopic quantum coherence and its screening transitions in two dimensions.

cond-mat.mes-hall↗

arXiv2026

Topological Surface Charge Detection via Active Capacitive Compensation: A Pathway to the 4D Quantum Hall Effect

The topological magnetoelectric effect (TME) in three-dimensional topological insulators (TIs), described by $\Delta P = \frac{e^2}{2h} N_{\rm Ch}^{(2)} \Delta B$, serves as a condensed-matter realization of the four-dimensional quantum Hall effect (4D QHE). In dual-gate axion-insulator devices, the TME-induced polarization yields a current $I_{\rm TME} \propto (C_{\rm total}/C_{\rm S})\,Q_{\rm 4D\mathrm{-}QHE}$, where the signal is suppressed by the capacitance ratio $C_{\rm total}/C_{\rm S}$. Here we propose an active compensation scheme that introduces a tunable negative capacitance $C_{\rm comp} \approx -C_{\rm gate}$ into the gate line, effectively canceling the gate dielectric capacitance and driving $C_{\rm total}/C_{\rm S} \to 1$. We validate the method using a quantum anomalous Hall (QAH) device, which shares the same surface-state physics as the axion insulator but permits direct charge measurement via a single gate, recovering over $95\%$ of the quantized charge signal from an initially half-attenuated state. This compensation method provides a robust means of resolving minute TME signals, offering a promising pathway toward direct measurements of the 4D QHE.

cond-mat.mes-hall↗

arXiv2026

The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context

In the world of Harry Potter, when Dumbledore's mind is overburdened, he extracts memories into a Pensieve to be revisited later. In the world of AI, while we possess the Pensieve-mature databases and retrieval systems, our models inexplicably lack the "wand" to operate it. They remain like a Dumbledore without agency, passively accepting a manually engineered context as their entire memory. This work finally places the wand in the model's hand. We introduce StateLM, a new class of foundation models endowed with an internal reasoning loop to manage their own state. We equip our model with a suite of memory tools, such as context pruning, document indexing, and note-taking, and train it to actively manage these tools. By learning to dynamically engineering its own context, our model breaks free from the architectural prison of a fixed window. Experiments across various model sizes demonstrate StateLM's effectiveness across diverse scenarios. On long-document QA tasks, StateLMs consistently outperform standard LLMs across all model scales; on the chat memory task, they achieve absolute accuracy improvements of 10% to 20% over standard LLMs. On the deep research task BrowseComp-Plus, the performance gap becomes even more pronounced: StateLM achieves up to 52% accuracy, whereas standard LLM counterparts struggle around 5%. Ultimately, our approach shifts LLMs from passive predictors to state-aware agents where reasoning becomes a stateful and manageable process.

cs.AI↗

arXiv2026

Free(): Learning to Forget in Malloc-Only Reasoning Models

Reasoning models enhance problem-solving by scaling test-time compute, yet they face a critical paradox: excessive thinking tokens often degrade performance rather than improve it. We attribute this to a fundamental architectural flaw: standard LLMs operate as "malloc-only" engines, continuously accumulating valid and redundant steps alike without a mechanism to prune obsolete information. To break this cycle, we propose Free()LM, a model that introduces an intrinsic self-forgetting capability via the Free-Module, a plug-and-play LoRA adapter. By iteratively switching between reasoning and cleaning modes, Free()LM dynamically identifies and prunes useless context chunks, maintaining a compact and noise-free state. Extensive experiments show that Free()LM provides consistent improvements across all model scales (8B to 685B). It achieves a 3.3% average improvement over top-tier reasoning baselines, even establishing a new SOTA on IMOanswerBench using DeepSeek V3.2-Speciale. Most notably, in long-horizon tasks where the standard Qwen3-235B-A22B model suffers a total collapse (0% accuracy), Free()LM restores performance to 50%. Our findings suggest that sustainable intelligence requires the freedom to forget as much as the power to think.

cs.AI↗

arXiv2025

Bottom-up Policy Optimization: Your Language Model Policy Secretly Contains Internal Policies

Existing reinforcement learning (RL) approaches treat large language models (LLMs) as a unified policy, overlooking their internal mechanisms. In this paper, we decompose the LLM-based policy into Internal Layer Policies and Internal Modular Policies via the Transformer's residual stream. Our entropy analysis of internal policy reveals distinct patterns: (1) universally, internal policies evolve from high-entropy exploration in early layers to deterministic refinement in the top layers; and (2) Qwen exhibits an explicit progressive reasoning structure, contrasting with the abrupt convergence in Llama. Furthermore, we discover that optimizing internal layers induces feature refinement, forcing lower layers to capture high-level reasoning representations early. Motivated by these findings, we propose Bottom-up Policy Optimization (BuPO), a novel RL paradigm that reconstructs the LLM's reasoning foundation from the bottom up by optimizing internal layers in early stages. Extensive experiments on complex reasoning benchmarks demonstrate the effectiveness of BuPO.

cs.LG↗

arXiv2025

DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains

Large Reasoning Models (LRMs) have demonstrated impressive capabilities but suffer from cognitive inefficiencies like "overthinking" simple problems and "underthinking" complex ones. While existing methods that use supervised fine-tuning (SFT) or reinforcement learning (RL) with token-length rewards can improve efficiency, they often do so at the cost of accuracy. This paper introduces DeepCompress, a novel framework that simultaneously enhances both the accuracy and efficiency of LRMs. We challenge the prevailing approach of consistently favoring shorter reasoning paths, showing that longer responses can contain a broader range of correct solutions for difficult problems. DeepCompress employs an adaptive length reward mechanism that dynamically classifies problems as "Simple" or "Hard" in real-time based on the model's evolving capability. It encourages shorter, more efficient reasoning for "Simple" problems while promoting longer, more exploratory thought chains for "Hard" problems. This dual-reward strategy enables the model to autonomously adjust its Chain-of-Thought (CoT) length, compressing reasoning for well-mastered problems and extending it for those it finds challenging. Experimental results on challenging mathematical benchmarks show that DeepCompress consistently outperforms baseline methods, achieving superior accuracy while significantly improving token efficiency.

cs.AI↗

arXiv2025

Investigating the relationship between the Weyl semimetal phase and the three-dimensional quantum Hall phase in ZrTe$_5$

The material ZrTe$_5$ exhibits distinct topological phases, including a Weyl semimetal phase, characterized by a chiral anomaly and in-plane Hall effect, and a three-dimensional quantum Hall phase. The relationship between these phases remains poorly understood. This work systematically explores their connection in ZrTe$_5$ through rotatable, pressure-dependent measurements. At ambient pressure, both phases are observed; the WSM phase requires strong electronic polarization, while the 3D QH phase appears when the characteristic resistivity peak temperature $T_p$ is approximately 90 K. Under applied pressure, the polarization diminishes, weakening the WSM phase and its associated nontrivial Hall signals. Concurrently, $T_p$ rises dramatically from 2 K at ambient pressure to 70 K at 2.2 GPa, approaching the expected regime for the 3D QH phase. These findings clarify the conditions underlying the WSM and 3D QH phases and suggest that exploring the 3D QH phase at even higher pressures is a promising direction for future research.

cond-mat.other↗

arXiv2025

Observation of Quantized Charge Accumulation in a Quantum Anomalous Hall System

The quantum anomalous Hall effect in magnetically doped topological insulators exhibits a quantized Hall conductance $\sigma_{xy} = e^2/h$ arising from the two-dimensional surface states. While conventional transport probes confirm this quantization, they remain insensitive to the field-induced surface charge accumulation as a direct manifestation of $\sigma_{xy}$. Here, we experimentally validate an out-of-plane capacitive method that directly detects this quantized charge accumulation in a quantum anomalous Hall system. Using Corbino and simple disk devices, we measure charge accumulation proportional to field variation $\Delta B$, with dissipation characterized by longitudinal conductance $\sigma_{xx}$ and frequency $f$. A quantitative dissipation model extracts the intrinsic quantized charge density $\eta_0 = (e^2/h)\Delta B$, which is confirmed through $f$- and $\sigma_{xx}$-dependent measurements. Under ultra-low dissipation ($\sigma_{xx} \approx 10^{-9}$ S), we directly resolve the fully quantized charge accumulation. This methodology establishes a direct charge-accumulation probe and provides a pathway toward detecting the topological magnetoelectric effect, a condensed matter manifestation of the four-dimensional quantum Hall effect.

cond-mat.mes-hall↗