arXiv subjects

Sangwoo Shin

Publications and source records attributed to Sangwoo Shin.

13 recordsLinked to original sources

arXiv2026

Articulated-Body Dynamics Network: Dynamics-Grounded Prior for Robot Learning

Recent work in reinforcement learning has shown that incorporating structural priors for articulated robots, such as link connectivity, into policy networks improves learning efficiency. However, dynamics properties, despite their fundamental role in determining how forces and motion propagate through the body, remain largely underexplored as an inductive bias for policy learning. To address this gap, we present the Articulated-Body Dynamics Network (ABD-Net), a novel graph neural network architecture grounded in the computational structure of forward dynamics. Specifically, we adapt the inertia propagation mechanism from the Articulated Body Algorithm, systematically aggregating inertial quantities from child to parent links in a tree-structured manner, while replacing physical quantities with learnable parameters. Embedding ABD-NET into the policy actor enables dynamics-informed representations that capture how actions propagate through the body, leading to efficient and robust policy learning. Through experiments with simulated humanoid, quadruped, and hopper robots, our approach demonstrates increased sample efficiency and generalization to dynamics shifts compared to transformer-based and GNN baselines. We further validate the learned policy on real Unitree G1 and Go2 robots, state-of-the-art humanoid and quadruped platforms, generating dynamic, versatile and robust locomotion behaviors through sim-to-real transfer with real-time inference.

cs.RO↗

arXiv2026

A Theoretical Analysis of Why Masked Diffusion Models Mitigate the Reversal Curse

Autoregressive language models (ARMs) suffer from the reversal curse: after learning ''$A$ is $B$,'' they often fail on the reverse query ''$B$ is $A$.'' Masked diffusion language models (MDMs) exhibit this failure in a much weaker form, but the underlying reason has remained unclear. A common explanation attributes this mitigation to their any-order masked training objective. However, observing ''$[\mathbf{M}]$ is $B$'' during training teaches recovery of $A$ from $B$ in one positional configuration, and does not by itself explain why the learned evidence should transfer to the reverse prompt ''$B$ is $[\mathbf{M}]$.'' We provide a theoretical analysis showing that this transfer arises from a parameter-level coupling between forward and reverse positional conditionals: shared Transformer parameters store token-pair evidence, while relative positional encodings route attention through queries and keys without changing the value-side evidence being retrieved. In a one-layer MDM, we prove that forward masked training strengthens evidence that is reusable in reverse queries, induces correlated forward--reverse attention routes, and yields a positively aligned shared-storage gradient component that decreases the reverse loss to first order. Controlled one-layer experiments and large-scale LLaDA/Dream experiments verify these signatures and show that they translate into improved reverse prediction.

cs.AI↗

arXiv2025

Information-Theoretic Discrete Diffusion

We present an information-theoretic framework for discrete diffusion models that yields principled estimators of log-likelihood using score-matching losses. Inspired by the I-MMSE identity for the Gaussian setup, we derive analogous results for the discrete setting. Specifically, we introduce the Information-Minimum Denoising Score Entropy (I-MDSE) relation, which links mutual information between data and its diffused version to the minimum denoising score entropy (DSE) loss. We extend this theory to masked diffusion and establish the Information-Minimum Denoising Cross-Entropy (I-MDCE) relation, connecting cross-entropy losses to mutual information in discrete masked processes. These results provide a time-integral decomposition of the log-likelihood of the data in terms of optimal score-based losses, showing that commonly used losses such as DSE and DCE are not merely variational bounds but tight and principled estimators of log-likelihood. The I-MDCE decomposition further enables practical extensions, including time-free formula, conditional likelihood estimation in prompt-response tasks, and coupled Monte Carlo estimation of likelihood ratios. Experiments on synthetic and real-world data confirm the accuracy, variance stability, and utility of our estimators. The code is publicly available at https://github.com/Dongjae0324/infodis.

cs.LG↗

arXiv2025

Stabilizing by steering: Enhancing bacterial motility by non-uniform diffusiophoresis

Bacteria often traverse confined spaces to perform critical functions in symbiosis, infection, drug delivery, and soil bioremediation. While the canonical run-and-tumble strategy enables exploration, its reliance on constant sensing and stochastic reorientation limits efficiency under confinement. We show that salt gradients can physically steer Pseudomonas putida by biasing their runs toward salt through asymmetric diffusiophoretic forces. These gradients impose a torque strong enough to overcome Brownian rotation, aligning cells along the gradient and producing straighter, more persistent motion. We further show that when toxic organic contaminants are present, salt gradients enhance bacterial dispersion toward them, demonstrating improved chemotactic transport. This work uncovers a previously unrecognized mechanism by which salt gradients direct bacterial motility, revealing a physical route to control microbial transport and colonization in complex environments.

physics.bio-ph↗

arXiv2025

A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models

Diffusion large language models (dLLMs) enable any-order generation, but this flexibility enlarges the attack surface: harmful spans may appear at arbitrary positions, and template-based prefilling attacks such as DIJA bypass response-level refusals. We introduce A2D (Any-Order, Any-Step Defense), a token-level alignment method that aligns dLLMs to emit an [EOS] refusal signal whenever harmful content arises. By aligning safety directly at the token-level under randomized masking, A2D achieves robustness to both any-decoding-order and any-step prefilling attacks under various conditions. It also enables real-time monitoring: dLLMs may begin a response but automatically terminate if unsafe continuation emerges. On safety benchmarks, A2D consistently prevents the generation of harmful outputs, slashing DIJA success rates from over 80% to near-zero (1.3% on LLaDA-8B-Instruct, 0.0% on Dream-v0-Instruct-7B), and thresholded [EOS] probabilities allow early rejection, yielding up to 19.3x faster safe termination.

cs.CL↗

arXiv2024

LLM-Based Offline Learning for Embodied Agents via Consistency-Guided Reward Ensemble

Employing large language models (LLMs) to enable embodied agents has become popular, yet it presents several limitations in practice. In this work, rather than using LLMs directly as agents, we explore their use as tools for embodied agent learning. Specifically, to train separate agents via offline reinforcement learning (RL), an LLM is used to provide dense reward feedback on individual actions in training datasets. In doing so, we present a consistency-guided reward ensemble framework (CoREN), designed for tackling difficulties in grounding LLM-generated estimates to the target environment domain. The framework employs an adaptive ensemble of spatio-temporally consistent rewards to derive domain-grounded rewards in the training datasets, thus enabling effective offline learning of embodied agents in different environment domains. Experiments with the VirtualHome benchmark demonstrate that CoREN significantly outperforms other offline RL agents, and it also achieves comparable performance to state-of-the-art LLM-based agents with 8B parameters, despite CoREN having only 117M parameters for the agent policy network and using LLMs only for training.

cs.AI↗

arXiv2024

Semantic Skill Grounding for Embodied Instruction-Following in Cross-Domain Environments

In embodied instruction-following (EIF), the integration of pretrained language models (LMs) as task planners emerges as a significant branch, where tasks are planned at the skill level by prompting LMs with pretrained skills and user instructions. However, grounding these pretrained skills in different domains remains challenging due to their intricate entanglement with the domain-specific knowledge. To address this challenge, we present a semantic skill grounding (SemGro) framework that leverages the hierarchical nature of semantic skills. SemGro recognizes the broad spectrum of these skills, ranging from short-horizon low-semantic skills that are universally applicable across domains to long-horizon rich-semantic skills that are highly specialized and tailored for particular domains. The framework employs an iterative skill decomposition approach, starting from the higher levels of semantic skill hierarchy and then moving downwards, so as to ground each planned skill to an executable level within the target domain. To do so, we use the reasoning capabilities of LMs for composing and decomposing semantic skills, as well as their multi-modal extension for assessing the skill feasibility in the target domain. Our experiments in the VirtualHome benchmark show the efficacy of SemGro in 300 cross-domain EIF scenarios.

cs.AI↗

arXiv2024

One-shot Imitation in a Non-Stationary Environment via Multi-Modal Skill

One-shot imitation is to learn a new task from a single demonstration, yet it is a challenging problem to adopt it for complex tasks with the high domain diversity inherent in a non-stationary environment. To tackle the problem, we explore the compositionality of complex tasks, and present a novel skill-based imitation learning framework enabling one-shot imitation and zero-shot adaptation; from a single demonstration for a complex unseen task, a semantic skill sequence is inferred and then each skill in the sequence is converted into an action sequence optimized for environmental hidden dynamics that can vary over time. Specifically, we leverage a vision-language model to learn a semantic skill set from offline video datasets, where each skill is represented on the vision-language embedding space, and adapt meta-learning with dynamics inference to enable zero-shot skill adaptation. We evaluate our framework with various one-shot imitation scenarios for extended multi-stage Meta-world tasks, showing its superiority in learning complex tasks, generalizing to dynamics changes, and extending to different demonstration conditions and modalities, compared to other baselines.

cs.AI↗

arXiv2024

SemTra: A Semantic Skill Translator for Cross-Domain Zero-Shot Policy Adaptation

This work explores the zero-shot adaptation capability of semantic skills, semantically interpretable experts' behavior patterns, in cross-domain settings, where a user input in interleaved multi-modal snippets can prompt a new long-horizon task for different domains. In these cross-domain settings, we present a semantic skill translator framework SemTra which utilizes a set of multi-modal models to extract skills from the snippets, and leverages the reasoning capabilities of a pretrained language model to adapt these extracted skills to the target domain. The framework employs a two-level hierarchy for adaptation: task adaptation and skill adaptation. During task adaptation, seq-to-seq translation by the language model transforms the extracted skills into a semantic skill sequence, which is tailored to fit the cross-domain contexts. Skill adaptation focuses on optimizing each semantic skill for the target domain context, through parametric instantiations that are facilitated by language prompting and contrastive learning-based context inferences. This hierarchical adaptation empowers the framework to not only infer a complex task specification in one-shot from the interleaved multi-modal snippets, but also adapt it to new domains with zero-shot learning abilities. We evaluate our framework with Meta-World, Franka Kitchen, RLBench, and CARLA environments. The results clarify the framework's superiority in performing long-horizon tasks and adapting to different domains, showing its broad applicability in practical use cases, such as cognitive robots interpreting abstract instructions and autonomous vehicles operating under varied configurations.

cs.AI↗

arXiv2023

Flow Driven Oil Recovery Enhanced with Structural Disjoining Pressure

Nanofluids have the potential to enhance oil recovery through the structural disjoining pressure, a pressure developed when nanoparticles concentrate at the three-phase contact line. A model microfluidic porous network is used to measure the percentage of oil displaced from this channel as the volume fraction of a Triton X-100 micelle nanofluid is varied from 0 - 30%. The percentage of oil displaced varies nearly linearly with micellar nanoparticle volume fraction starting with 39% using deionized water and 89% using a volume fraction of 30%. While the trend is clear, significant variability between experiments was observed for a fixed nanofluid volume fraction. This indicates that surface energy heterogeneity is important for the nanofluid oil displacement performance.

physics.flu-dyn↗

arXiv2023

Shape- and orientation-dependent diffusiophoresis of colloidal ellipsoids

We present the diffusiophoresis of ellipsoidal particles induced by ionic solute gradients. Contrary to the common expectation that diffusiophoresis is shape independent, here we show experimentally that this assumption breaks down when the thin Debye layer approximation is relaxed. By tracking the translation and rotation of various ellipsoids, we find that the phoretic mobility of ellipsoids is sensitive to the eccentricity and the orientation of the ellipsoid relative to the imposed solute gradient, and can further lead to nonmonotonic behavior under strong confinement. We show that such a shape- and orientation-dependent diffusiophoresis of colloidal ellipsoids can be easily captured by modifying theories for spheres.

cond-mat.soft↗

arXiv2023

Viscosity measurements of glycerol in a parallel-plate rheometer exposed to atmosphere

Glycerol is a hygroscopic fluid that spontaneously absorbs water vapor from the atmosphere. For applications involving glycerol, care must be taken to avoid exposure to humidity, since its viscosity decreases quickly as water is absorbed. We report experimental measurements of the viscosity of glycerol in a parallel-plate rheometer where the outer interface is exposed to atmosphere. The measurements decrease with time as water is absorbed from the atmosphere and transported throughout the glycerol via diffusion and advection. Measured viscosities drop faster at higher relative humidities, confirming the role of hygroscopicity on the transient viscosities. The rate of viscosity decrease shows a non-monotonic relationship with the rheometer gap height. This behavior is explained by considering the transition from diffusion-dominated transport in the narrow gap regime to the large gap regime where transport is dominated by inertia-driven secondary flows. Numerical simulations of the water absorption and transport confirm this non-monotonic behavior. The experimental viscosity measurements show unexpectedly fast decreases at very small gap heights, violating the parallel-plate, axisymmetric model. We propose that this drop-off may be due to misalignment in the rheometer that becomes non-negligible for small gaps. Theoretical considerations show that secondary flows in a misaligned rheometer dominate the typical secondary inertial flows in parallel-plate rheometers at small gaps. Finally, simulations in a misaligned parallel-plate system demonstrate the same sharp drop-off in viscosity measurements at small gap heights. This modeling can be used to estimate the gap height where misalignment effects dominate the transient glycerol viscosity measurements.

physics.flu-dyn↗

arXiv2017

Diffusiophoretic manipulation of particles in a drop deposited on a hydrogel

We report an experimental study on the manipulation of colloidal particles in a drop sitting on a hydrogel. The manipulation is achieved by diffusiophoresis, which describes a directed motion of particles induced by solute gradients. By letting the solute concentrations for the drop and the hydrogel be different, we control the motion of particles in a stable suspension, which is otherwise difficult to achieve. We show that diffusiophoresis can cause the particles to move either toward or away from the liquid-air interface depending on the direction of the solute gradient and the surface charge of the particles. We measure the particle adsorption experimentally and rationalize the results with a one-dimensional numerical model. We show that diffusiophoretic motion is significant at the lengthscale of a drop deposited on a hydrogel, which suggests a simple method for the deposition of particles on hydrogels.

cond-mat.soft↗