arXiv subjects

Liang Tan

Publications and source records attributed to Liang Tan.

At least 19 recordsLinked to original sources

arXiv2026

WorldExam: Benchmarking World Models from Apparent Appearance to Inherent Reactivity

Controllable video generation models are increasingly being developed as world models. Accordingly, evaluating them in this role extends beyond the apparent appearance of generated videos to the inherent reactivity of the worlds they depict: the ability to infer from the scene state how the world should react and to generate plausible consequences not explicitly described in the input. Yet existing benchmarks mainly assess visual quality or explicit instruction fulfillment by checking whether requested actions and interaction outcomes are realized, leaving inherent reactivity underexamined. We introduce WorldExam, a hierarchical diagnostic benchmark spanning four levels: Visual Quality, Control Adherence, Spatial Consistency, and World Reactivity. It comprises 1,474 cases across eight dedicated tasks and supports unified evaluation of camera-, action-, and language-driven model paradigms. The World Reactivity level evaluates scene-conditioned reactions and goal-directed behaviors beyond what is explicitly specified in the input. Evaluation of 20 representative models reveals a clear capability split. Camera-driven models excel at camera control, but their interfaces do not support dynamic interaction; action-driven models control subjects more precisely but often leave the world unresponsive; and language-driven models perform better on interaction but follow complex controls less faithfully. No model combines broad task coverage with consistently strong performance, showing that high visual quality and explicit instruction fulfillment do not guarantee inherent reactivity.

cs.CV↗

arXiv2025

Mesh Augmentation of LoRaWAN-based IoT Networks

LoRaWAN is a leading standard and technology for low-power, long-range Internet-of-Things (IoT) communications. However, its single-hop architecture results in limited effective range and excessive power consumption for end devices, especially when deployed in large, remote and RF-challenged environments. Existing solutions are either incompatible with LoRaWAN, or limit relaying to a single hop. We present LIMA, a protocol for augmenting an existing or new LoRaWAN deployment with a mesh network of LIMA Routers. LIMA increases the effective coverage range well beyond the maximum LoRa range via multi-hopping, and significantly reduces the energy consumed by end-devices. LIMA requires no changes to the end-device, the servers or the LoRaWAN standard. LIMA builds routes using reverse path forwarding, tunnels LoRaWAN messages over LIMA, provides transparent extension of the existing Adaptive Data Rate (ADR), and suppresses duplicate forwarding if the device is directly reachable from the Gateway. Simulations using Network Simulator 3 (ns-3) show that LIMA increases the delivery rate, scalability, ED energy consumption by up to 5x, 8x and 12.6x respectively, and reduces latency by up to 2.3x. Table-top and outdoor testing with a prototype constructed using a commercial gateway as a starting point confirm that LIMA can be successfully deployed within an existing LoRaWAN system, and can provide range and energy gains transparently.

cs.NI↗

arXiv2025

Prompt Curriculum Learning for Efficient LLM Post-Training

We introduce Prompt Curriculum Learning (PCL), a lightweight reinforcement learning (RL) algorithm that selects intermediate-difficulty prompts using a learned value model to post-train language models. Since post-training LLMs via RL remains sensitive to batching and prompt selection strategies, we first conduct a series of systematic experiments where we (1) determine the optimal training batch size that balances generation efficiency and gradient quality and (2) establish the importance of focusing on prompts of intermediate difficulty for the policy. We build upon these results to design PCL, which identifies prompts of intermediate difficulty for the current policy in an on-policy manner by using a value model that is concurrently updated based on the current policy. By focusing on informative prompts that yield high effective ratios, PCL achieves either the highest performance or requires significantly less time to reach comparable performance to its counterparts. Compared to rollout-based filtering methods, PCL avoids costly rollouts and achieves $12.1\times$ and $16.9\times$ faster speed on identifying intermediate-difficulty prompts when training on MATH and DeepScaleR, respectively. We further demonstrate that our value model accurately predicts prompt difficulty and allows PCL to focus on progressively more challenging prompts during RL. Our results present a new methodology that delivers improved tradeoff between upper-bound performance and efficiency for reasoning-focused RL.

cs.LG↗

arXiv2025

Your thoughts tell who you are: Characterize the reasoning patterns of LRMs

Current comparisons of large reasoning models (LRMs) focus on macro-level statistics such as task accuracy or reasoning length. Whether different LRMs reason differently remains an open question. To address this gap, we introduce the LLM-proposed Open Taxonomy (LOT), a classification method that uses a generative language model to compare reasoning traces from two LRMs and articulate their distinctive features in words. LOT then models how these features predict the source LRM of a reasoning trace based on their empirical distributions across LRM outputs. Iterating this process over a dataset of reasoning traces yields a human-readable taxonomy that characterizes how models think. We apply LOT to compare the reasoning of 12 open-source LRMs on tasks in math, science, and coding. LOT identifies systematic differences in their thoughts, achieving 80-100% accuracy in distinguishing reasoning traces from LRMs that differ in scale, base model family, or objective domain. Beyond classification, LOT's natural-language taxonomy provides qualitative explanations of how LRMs think differently. Finally, in a case study, we link the reasoning differences to performance: aligning the reasoning style of smaller Qwen3 models with that of the largest Qwen3 during test time improves their accuracy on GPQA by 3.3-5.7%.

cs.CL↗

arXiv2025

Probing Light Dark Matter with Cosmic Gravitational Focusing

We investigate the possibility of using the cosmic gravitational focusing (CGF) to probe the minor light dark matter (DM) component whose mass is in the range of $(0.1 \sim 100)$\,eV. Being a purely gravitational effect, the CGF offers a mode-independent probe that is complementary to the existing ways such as Lyman-$\alpha$ and $\Delta N_{\rm eff}$. Such effect finally leads to a dipole density distribution that would affect the galaxy formation and hence can be reconstructed with galaxy surveys such as DESI. Both the free-streaming and clustering limits have been studied with analytical formulas while the region in between is bridged with interpolation. We show the projected sensitivity at DESI with the typical phase space distribution of a freeze-in DM scenario as illustration.

hep-ph↗

arXiv2025

Active noise cancellation in ultra-low field MRI: distinct strategies for different channels

Ultra-low field magnetic resonance imaging(ULF-MRI) systems operating in open environments are highly susceptible to composite electromagnetic interference(EMI). Different imaging channels respond non-uniformly to EMI owing to their distinct coupling characteristics. Here, we investigate channel-specific interference pathways in a permanent-magnet-based low-field MRI system and show that saddle coils are intrinsically more vulnerable to transverse EMI components than solenoidal coils. To mitigate these heterogeneous coupling effects, we propose a dual-stage suppression strategy that combines front-end spatial-domain inverse field reconstruction with back-end channel-adaptive active noise cancellation. Experiments demonstrate that this approach suppresses EMI by more than 80%, substantially improves inter-channel signal-to-noise ratio(SNR) consistency, and enhances the fused-image SNR by 24%. These findings elucidate the channel-dependent nature of EMI coupling and establish targeted mitigation strategies, providing both a theoretical basis and practical guidance for noise suppression in future array-coil ULF-MRI systems.

eess.SP↗

arXiv2025

ASTRO: Teaching Language Models to Reason by Reflecting and Backtracking In-Context

We introduce ASTRO, the "Autoregressive Search-Taught Reasoner", a framework for training language models to reason like search algorithms, explicitly leveraging self-reflection, backtracking, and exploration in their outputs. Recently, training large language models (LLMs) via reinforcement learning (RL) has led to the advent of reasoning models with greatly enhanced reasoning capabilities. Open-source replications of reasoning models, while successful, build upon models that already exhibit strong reasoning capabilities along with search behavior observed even before RL. As a result, it is yet unclear how to boost the reasoning capabilities of other non-reasoner models including Llama 3. ASTRO teaches such models to internalize structured search behavior through a synthetic dataset derived from Monte Carlo Tree Search (MCTS) over mathematical problem-solving trajectories. By converting search traces into natural language chain-of-thoughts that capture both successes and recoveries from failure, ASTRO bootstraps models with a rich prior for exploration during RL. We finetune our models on these search-derived traces and further improve performance via RL with verifiable rewards. We apply ASTRO to the Llama 3 family of models and achieve absolute performance gains of 16.0% on MATH-500, 26.9% on AMC 2023, and 20.0% on AIME 2024, especially improving upon challenging problems that require iterative correction. Our results demonstrate that search-inspired training offers a principled way to instill robust reasoning capabilities into open LLMs.

cs.AI↗

arXiv2025

Boosting LLM Reasoning via Spontaneous Self-Correction

While large language models (LLMs) have demonstrated remarkable success on a broad range of tasks, math reasoning remains a challenging one. One of the approaches for improving math reasoning is self-correction, which designs self-improving loops to let the model correct its own mistakes. However, existing self-correction approaches treat corrections as standalone post-generation refinements, relying on extra prompt and system designs to elicit self-corrections, instead of performing real-time, spontaneous self-corrections in a single pass. To address this, we propose SPOC, a spontaneous self-correction approach that enables LLMs to generate interleaved solutions and verifications in a single inference pass, with generation dynamically terminated based on verification outcomes, thereby effectively scaling inference time compute. SPOC considers a multi-agent perspective by assigning dual roles -- solution proposer and verifier -- to the same model. We adopt a simple yet effective approach to generate synthetic data for fine-tuning, enabling the model to develop capabilities for self-verification and multi-agent collaboration. We further improve its solution proposal and verification accuracy through online reinforcement learning. Experiments on mathematical reasoning benchmarks show that SPOC significantly improves performance. Notably, SPOC boosts the accuracy of Llama-3.1-8B and 70B Instruct models, achieving gains of 8.8% and 11.6% on MATH500, 10.0% and 20.0% on AMC23, and 3.3% and 6.7% on AIME24, respectively.

cs.AI↗

arXiv2025

LlamaRL: A Distributed Asynchronous Reinforcement Learning Framework for Efficient Large-scale LLM Training

Reinforcement Learning (RL) has become the most effective post-training approach for improving the capabilities of Large Language Models (LLMs). In practice, because of the high demands on latency and memory, it is particularly challenging to develop an efficient RL framework that reliably manages policy models with hundreds to thousands of billions of parameters. In this paper, we present LlamaRL, a fully distributed, asynchronous RL framework optimized for efficient training of large-scale LLMs with various model sizes (8B, 70B, and 405B parameters) on GPU clusters ranging from a handful to thousands of devices. LlamaRL introduces a streamlined, single-controller architecture built entirely on native PyTorch, enabling modularity, ease of use, and seamless scalability to thousands of GPUs. We also provide a theoretical analysis of LlamaRL's efficiency, including a formal proof that its asynchronous design leads to strict RL speed-up. Empirically during the Llama 3 post-training, by leveraging best practices such as colocated model offloading, asynchronous off-policy training, and distributed direct memory access for weight synchronization, LlamaRL achieves significant efficiency gains -- up to 10.7x speed-up compared to DeepSpeed-Chat-like systems on a 405B-parameter policy model. Furthermore, the efficiency advantage continues to grow with increasing model scale, demonstrating the framework's suitability for future large-scale RL training.

cs.LG↗

arXiv2024

A Systematic Examination of Preference Learning through the Lens of Instruction-Following

Preference learning is a widely adopted post-training technique that aligns large language models (LLMs) to human preferences and improves specific downstream task capabilities. In this work we systematically investigate how specific attributes of preference datasets affect the alignment and downstream performance of LLMs in instruction-following tasks. We use a novel synthetic data generation pipeline to generate 48,000 unique instruction-following prompts with combinations of 23 verifiable constraints that enable fine-grained and automated quality assessments of model responses. With our synthetic prompts, we use two preference dataset curation methods - rejection sampling (RS) and Monte Carlo Tree Search (MCTS) - to obtain pairs of (chosen, rejected) responses. Then, we perform experiments investigating the effects of (1) the presence of shared prefixes between the chosen and rejected responses, (2) the contrast and quality of the chosen, rejected responses and (3) the complexity of the training prompts. Our experiments reveal that shared prefixes in preference pairs, as generated by MCTS, provide marginal but consistent improvements and greater stability across challenging training configurations. High-contrast preference pairs generally outperform low-contrast pairs; however, combining both often yields the best performance by balancing diversity and learning efficiency. Additionally, training on prompts of moderate difficulty leads to better generalization across tasks, even for more complex evaluation scenarios, compared to overly challenging prompts. Our findings provide actionable insights into optimizing preference data curation for instruction-following tasks, offering a scalable and effective framework for enhancing LLM training and alignment.

cs.CL↗

arXiv2024

Self-Generated Critiques Boost Reward Modeling for Language Models

Reward modeling is crucial for aligning large language models (LLMs) with human preferences, especially in reinforcement learning from human feedback (RLHF). However, current reward models mainly produce scalar scores and struggle to incorporate critiques in a natural language format. We hypothesize that predicting both critiques and the scalar reward would improve reward modeling ability. Motivated by this, we propose Critic-RM, a framework that improves reward models using self-generated critiques without extra supervision. Critic-RM employs a two-stage process: generating and filtering high-quality critiques, followed by joint fine-tuning on reward prediction and critique generation. Experiments across benchmarks show that Critic-RM improves reward modeling accuracy by 3.7%-7.3% compared to standard reward models and LLM judges, demonstrating strong performance and data efficiency. Additional studies further validate the effectiveness of generated critiques in rectifying flawed reasoning steps with 2.5%-3.2% gains in improving reasoning accuracy.

cs.CL↗

arXiv2024

High sensitivity pressure and temperature quantum sensing in organic crystals

The inherent sensitivity of quantum sensors to their physical environment can make them good reporters of parameters such as temperature, pressure, strain, and electric fields. Here, we present a molecular platform for pressure (P) and temperature (T) sensing using para-terphenyl crystals doped with pentacene. We leverage the optically detected magnetic resonance (ODMR) of the photoexcited triplet electron in the pentacene molecule, that serves as a sensitive probe for lattice changes in the host para-terphenyl due to pressure or temperature variations. We observe maximal ODMR frequency variations of df/dP=1.8 MHz/bar and df/dT=247 kHz/K, which are over 1,200 times and three times greater, respectively, than those seen in nitrogen-vacancy centers in diamond. This results in a >85-fold improvement in pressure sensitivity over best previously reported. The larger variation reflects the weaker nature of the para-terphenyl lattice, with first-principles DFT calculations indicating that even picometer-level shifts in the molecular orbitals due to P, T changes are measurable. The platform offers additional advantages including high levels of sensor doping, narrow ODMR linewidths and high contrasts, and ease of deployment, leveraging the ability for large single crystals at low cost. Overall, this work paves the way for low-cost, optically-interrogated pressure and temperature sensors and lays the foundation for even more versatile sensors enabled by synthetic tunability in designer molecular systems.

quant-ph↗

arXiv2024

Law of the Weakest Link: Cross Capabilities of Large Language Models

The development and evaluation of Large Language Models (LLMs) have largely focused on individual capabilities. However, this overlooks the intersection of multiple abilities across different types of expertise that are often required for real-world tasks, which we term cross capabilities. To systematically explore this concept, we first define seven core individual capabilities and then pair them to form seven common cross capabilities, each supported by a manually constructed taxonomy. Building on these definitions, we introduce CrossEval, a benchmark comprising 1,400 human-annotated prompts, with 100 prompts for each individual and cross capability. To ensure reliable evaluation, we involve expert annotators to assess 4,200 model responses, gathering 8,400 human ratings with detailed explanations to serve as reference examples. Our findings reveal that, in both static evaluations and attempts to enhance specific abilities, current LLMs consistently exhibit the "Law of the Weakest Link," where cross-capability performance is significantly constrained by the weakest component. Specifically, across 58 cross-capability scores from 17 models, 38 scores are lower than all individual capabilities, while 20 fall between strong and weak, but closer to the weaker ability. These results highlight the under-performance of LLMs in cross-capability tasks, making the identification and improvement of the weakest capabilities a critical priority for future research to optimize performance in complex, multi-dimensional scenarios.

cs.AI↗

arXiv2024

Identifying Neutrino Mass Ordering with Cosmic Gravitational Focusing

The cosmic gravitational focusing (CGF) of relic neutrinos can provide an independent measurement of the absolute neutrino masses $m_i$ with fourth-power dependence ($m^4_i$). We demonstrate in this paper for the first time how this can help identifying the neutrino mass ordering, using the fact that total mass falling below the inverted ordering threshold allows the discrimination of the inverted ordering. Upon incorporating the projected CGF sensitivity at DESI, the preference for the normal ordering with a prior $\sum m_i > 0.059\,{\rm eV}$ would increase from the original 89.9\% of the existing matter clustering method with the DESI analysis to 98.2\% while the inverted ordering is further disfavored from 10.1\% to 1.8\%. We also show how this can affect the prospects of the neutrinoless double beta decay and single beta decay measurements.

hep-ph↗

arXiv2024

Gemini Dark Matter

The $S_8/\sigma_8$ tension in the large scale structure can be explained by decaying dark matter with an almost degenerate spectrum and small enough decay width. Here we propose the Gemini dark matter model, which contains a heavy mother particle $\chi_3$ and two twins $\chi_{1/2}$ which are almost degenerate in mass and are produced at the same time. The dark sector is charged under the same Froggatt-Nielsen symmetry that can explain the hierarchy of the Standard model Yukawa couplings. The slightly heavier $\chi_2$ decays into $\chi_1$ and the axionic component of the flavon, which washes out the small scale structure and resolves $S_8/\sigma_8$ tension. We present the production mechanism of Gemini dark matter and viable parameter regions. We find that despite the preferred dark matter mass being $\mathcal{O}(1)$--$\mathcal{O}(100)$ keV, they constitute cold dark matter. The Gemini dark matter model predicts an abundance of dark radiation that will be probed in future measurements of the cosmic microwave background.

hep-ph↗

arXiv2023

Neutrino Mass Measurement with Cosmic Gravitational Focusing

We thoroughly explore the cosmic gravitational focusing of cosmic neutrino fluid (C$\nu$F) by dark matter (DM) halo using both general relativity for a point source of gravitational potential and Boltzmann equations for continuous overdensities. Derived in the general way for both relativistic and non-relativistic neutrinos, our results show that the effect has fourth power dependence on the neutrino mass and temperature. With nonlinear mass dependence which is different from the cosmic microwave background (CMB) and large scale structure (LSS) observations, the cosmic gravitational focusing can provide an independent cosmological way of measuring the neutrino mass and ordering. We take DESI as an example to illustrate that the projected sensitivity as well as its synergy with existing terrestrial neutrino oscillation experiments and other cosmological observations can significantly improve the neutrino mass measurement.

hep-ph↗

arXiv2023

Audiobox: Unified Audio Generation with Natural Language Prompts

Audio is an essential part of our life, but creating it often requires expertise and is time-consuming. Research communities have made great progress over the past year advancing the performance of large scale audio generative models for a single modality (speech, sound, or music) through adopting more powerful generative models and scaling data. However, these models lack controllability in several aspects: speech generation models cannot synthesize novel styles based on text description and are limited on domain coverage such as outdoor environments; sound generation models only provide coarse-grained control based on descriptions like "a person speaking" and would only generate mumbling human voices. This paper presents Audiobox, a unified model based on flow-matching that is capable of generating various audio modalities. We design description-based and example-based prompting to enhance controllability and unify speech and sound generation paradigms. We allow transcript, vocal, and other audio styles to be controlled independently when generating speech. To improve model generalization with limited labels, we adapt a self-supervised infilling objective to pre-train on large quantities of unlabeled audio. Audiobox sets new benchmarks on speech and sound generation (0.745 similarity on Librispeech for zero-shot TTS; 0.77 FAD on AudioCaps for text-to-sound) and unlocks new methods for generating audio with novel vocal and acoustic styles. We further integrate Bespoke Solvers, which speeds up generation by over 25 times compared to the default ODE solver for flow-matching, without loss of performance on several tasks. Our demo is available at https://audiobox.metademolab.com/

cs.SD↗

arXiv2023

Ultraheavy Atomic Dark Matter Freeze-Out through Rearrangement

Atomic dark matter is usually considered to be produced asymmetrically in the early Universe. In this work, we first propose that the symmetric atomic dark matter can be thermally produced through the freeze-out mechanism. The dominant atom anti-atom annihilation channel is the atomic rearrangement. It has a geometrical cross section much larger than that of elementary fermions. After the atomic formation, this annihilation process further depletes dark matter particles and finally freezes out. To give the observed dark matter relic, the dark atoms are naturally ultraheavy, ranging from $10^6$ to $10^{10} \,\mathrm{GeV}$.

hep-ph↗