EvoDuet: Bilevel Co-Evolution of Web Searching and Task Solving for Scientific Discovery

EvoDuet co-evolves solutions and web queries via a retrieval gate to boost LLM discovery gains up to 82.3% across optimization tasks.

Young-Jun Lee, Jinheon Baek, Soyeong Jeong, Minki Kang and 4 more

Published Sep 30, 2026 · 0 citations · ▲ 109 on Hugging Face · Code ★ 4

Beyond the Current Scene: Event-Referential Grasping with Active View Selection

BeyondSCe enables zero-shot event-referential grasping via active view selection, achieving 76% and 77% success on visible and occluded targets versus 40% and 55% baselines.

Hyunjoon Lee, Haebeom Jung, Eunsung Cha, Daeun Lee and 3 more

Published Sep 30, 2026 · 0 citations · ▲ 51 on Hugging Face · Code ★ 8

Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLMs

HeteroFold enables prefill-free cross-family KV cache transfer between frozen heterogeneous LLM agents, accelerating 32K context transfer up to 10.7x while matching text-based multi-agent performance.

Vincent-Daniel Yun, Woosang Lim, Haneul Yoo, Sungjoo Yoo and 2 more

Published Sep 26, 2026 · 0 citations · ▲ 95 on Hugging Face · Code ★ 1

SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation

JungHun Oh, Sungyong Baik, Kyoung Mu Lee

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Anatomy-aware Spatio-Temporal Modeling for Echocardiography Segmentation

Seok-Hwan Oh, Guil Jung, Myeong-Gee Kim, Young-Min Kim and 4 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Kernel Value Regression in Offline Reinforcement Learning

Jongyeon Lee, Jaehyoung Jeon, Kim Haneol, Myungjoo Kang

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Visual Grounding First, Multimodal In-context Learning Follows

Minhyuk Seo, Minjae Lee, Chaeeun Lee, Wei Lin and 3 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Heads That Write, Not Just Point: Image Retrieval Heads in Vision-Language Models

Junsung Park, Uiwon Hwang, Donghun Kang, Yeongtak Oh and 5 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

MM-SCALE: Evaluating Evidence-Grounded Moral Judgment in Vision-Language Models

Eunkyu Park, Wesley Deng, Cheyon Jin, Matheus Kunzler Maldaner and 7 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Values Are Not Single Labels: Distributional Value Profiling Across Groups and Contexts

Woojin Kim, Jaeyoung Do

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs

Hoigi Seo, Byung Hyun Lee, Minjun Kim, Dohyun Mah and 2 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

FLINT: Coupling Proximal Initialization and Bounded Stochasticity for Flow-Matching Inverse Problems

Junseo Bang, daewon choi, Dong Ju Mun, Se Young Chun

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

MINT: Meeting-time INdicators for Truncation in Multi-Step Off-Policy RL

Seungyub Han, Taehyun Cho, Dohyeong Kim, Kyungjae Lee and 1 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Cooperative Multi-Agent Reinforcement Learning via Epigraph-Form Guided Exploration

Sungil Son, Hoseong Jung, Dahyun Oh, H. Jin Kim

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Bilinear Matching Bandits

Wooseong Cho, Min-hwan Oh

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Sharper Regret Bounds for Shampoo

Dahngoon Kim, Min-hwan Oh

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

A Unified Semismooth Newton Approach to Multitask and Multivariate Square-Root Lasso Problems

Dongwon Kim, Taehyoung Kim, Sungdong Lee, Joong-Ho (Johann) Won

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

SNACK: A Sequential Notation Framework for Probabilistic Graph Generation

Hohyun Kim, Hyesung Kim, Min-hwan Oh, Seunggeun Lee

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Stochastic Matching Bandits with Rare Optimization Updates

Jung-hun Kim, Min-hwan Oh

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

Predicting Human-Gain Curves under Local Proxy Optimization from Repeated Ratings

Moonwon Choi, Kisung Nam, Seunggeun Lee

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Decomposing Conformal Uncertainty: Calibration- and Instance-Driven Feature Attribution

Sangyeon Cho, Minyoung Cho, Jungsoo Kim, Sujeong Oh and 1 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Distributionally Robust Domain Randomization with Learned Risk-Sensitive Dynamics Samplers

Sukchul Jeong, Insoon Yang

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

From Matching to Reasoning: Query-Aware Long Video Summarization

Mingu Kang, Sumin Kim, Hyunjin Lee, Sungmin Yang and 2 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Fractional Power-of-Two Quantization for Efficient and Effective Multiplier-Free LLM Inference

Sunghyun Wee, Geunjae Choi, Hyeonjin Kim, Suyoung Kim and 1 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Auditing Capsule Vision 2024: Within-Split Train-to-Validation Re-Exposure and a Kvasir-Channel Sensitivity Diagnostic

Eunseob Choi, Kyeonghun Kim, Hyuk-Jae Lee, Nam-Joon Kim

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Auditing Correlated Failures in Frozen-Feature Pretrained-Encoder Pools for Medical Segmentation

Eunseob Choi, Kyeonghun Kim, Hyuk-Jae Lee, Nam-Joon Kim

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

DictLLM: Post-training Compression of Large Language Model with Dictionary Kernels

Jinho PARK, Se Young Chun, Mingoo Seok

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

ManiFusion: Unlocking High-Throughput Generation via Superposition in Manifold Space

Minkyu Kim, Baekseung Kim, Junhoo Lee, Jangho Kim and 1 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Linear Contextual Bandits with Quasi-Optimism

Min-hwan Oh, Harin Lee

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Constrained Modulatory Reservoirs for Context-Dependent Computation

Sejeung Choi, Minji Jung, Jinyeong Park, Taek D Chung

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Shared-Noise Mechanisms for Verifiable Differentially Private Counting

Sehyeon Park, Chenglin Fan

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

From Static Geometry to Dynamical Singularity: Detecting Memorization in Diffusion Models via Score Evolution

Jeonseong Kim, Chenglin Fan

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

HetCCL: Efficient LLM Training on Heterogeneous Vendor GPUs

Heehoon Kim, Chris Jaehwan Lee, Taejeoung Kim, Jongwon Park and 5 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Risk-Sensitive Deep Optimal Stopping

Yechan Jeong, Youngjoon Hong

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Predicting Nothing Beats SAM 3: Revisiting Evaluation in Video Object Segmentation

Jihwan Hong, Woohyeon Park, Jaeik Kim, Jaeyoung Do

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Discrete Neural Interlingua for Ontology-Agnostic EHR Modeling

Tong Ding, Caiwei Tian, Dongmin Bang, Ming Yang Lu and 5 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Variance-Adaptive Optimal Algorithm for Reinforcement Learning with MNL Function Approximation

Wonyoung Kim, Garud Iyengar, Assaf Zeevi, Min-hwan Oh

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

WebArena-Pro: A Heterogeneous, Multimodal, Reproducible Benchmark for Web Agents

Imene Kerboua, Fatemeh Pesaran Zadeh, Xing Han Lu, Weijian Qi and 18 more

Paris Poster Session 4, Thu, Dec 10, 5:30 PM–7:30 PM, Paris Poster Hall · Published 2026

PDE-PFN: Prior-Data Fitted Neural PDE Solver

Jaehyeon Park, Mingu Kang, Dongseok Lee, Woojin Cho and 4 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Clean Data Can Still Carry Backdoors: Support-Persistent Backdoors for Model Reuse

Junhoo Lee, Baekseung Kim, Seungyeon Kim, Nojun Kwak

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Discriminative Score Function: Turning Pretrained Models into Functional Generative Priors

Junhoo Lee, Hyeonjin Kim, Sangbum Han, Nojun Kwak

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Generalized Influence Functions for Better Model Change Estimates

Hyeonsu Lyu, Jonggyu Jang, Sehyun Ryu, Hyun Jong Yang

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

LOCO: Local Light-Aware Object Compositing with Spatially Varying Illumination-Augmented Data

Jinseo Jeong, Hyunsoo Kim, Junseo Koo, Junhyeog Yun and 2 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

OntoPlan: An Ontology-Grounded Scene Representation and Agentic Framework for Scalable Robot Task Planning

Hyeongwoo Nam, Woongje Cho, Juwon Kim, Jongeun Choi

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

AEGIS: Almost Surely Safe Offline Reinforcement Learning

Junseo Lee, Hyeokjin Kwon, Songhwai Oh

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Structured Masked Diffusion for Joint Multiuser Decoding

CIDER is a masked-diffusion multiuser decoder using demixing and parity-aware propagation to outperform joint belief propagation by 6-100x in speed with matching error rates.

Taekyun Lee, Jiyoung Yun, Jeffrey Andrews, Hyeji Kim

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

Logistic Bandits with $\tilde{O}(\sqrt{dT})$ Regret without Context Diversity Assumptions

SupSplitLog achieves near-optimal logistic bandit regret without context diversity assumptions via sample splitting and Newton-type corrections.

Seoungbin Bae, Dabeen Lee

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Probabilistic Signature Inversion: Learning Conditional Distributions from Truncated Signatures

Truncated signature inversion is reframed as learning conditional path distributions via signature-conditioned flow matching, with derived Bayes error baselines and validated reconstruction on real data.

Junoh Kang, Kiseop Lee, Bohyung Han

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Reward Inflation: A Healthy Stimulus for Reinforcement Learning

Gradually scaling rewards during training accelerates RL adaptation by upweighting recent transitions, sustaining gradients to prevent neuron dormancy, and improving performance across diverse tasks.

Ganghun Lee, Minji Kim, Minsu Lee, Byoung-Tak Zhang

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

FiRe: Fine-grained Multimodal Reasoning for Enhanced Image Generation

FiRe improves image generation via fine-grained multimodal reasoning that decomposes prompts, self-checks visual requirements, and applies localized refinement, with FiRe-GRPO providing step-level reinforcement learning rewards.

Yongjin Kim, Yoonjin Oh, Ye Rin Kim, Hyomin Kim and 4 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 2

FoMo: Forking Moment in Generative Trajectory as a Perceptual Distance

Diffusion trajectory forking moments generate automatic perceptual distance labels that train reference-based image quality assessment metrics outperforming human-annotated datasets.

Jaihyun Lew, Mingi Jung, Minjun Park, Wooseok Song and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 24 on Hugging Face · Code ★ 11

The PokeAgent Challenge: Competitive and Long-Context Learning at Scale

PokeAgent is a large-scale Pokémon benchmark with battling and speedrunning tracks that expose major gaps between LLMs, RL agents, and human experts.

Seth Karten, Jake Grigsby, Tersoo Upaa, Junik Bae and 27 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 12 on Hugging Face

CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection

CompactAttention accelerates chunked prefill via block-union KV selection that builds minimal per-group block tables to eliminate KV copy overhead, achieving up to 2.72x attention speedup with near-dense accuracy on 128K contexts.

Jiwon Song, Dongwon Jo, Beomseok Kang, jae-joon kim

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 12 on Hugging Face · Code ★ 7

A Matched-Budget Audit Framework for Recaptioned Image-Text Supervision Distributions

A matched-budget audit framework profiles recaptioned image-text distributions via five axes and controllable basic units, showing released captions increase supported units by 3.39 to 6.36 and revealing a long-vs-dense frontier.

Giyeong Oh, Junghun Park, Yuhan Bae, Youngjae Yu

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Scenes as Objects, Not Primitives : Instance-Structured 3D Tokenization from Unposed Views

Feed-forward framework decomposes scenes into instance-structured 3D token groups from unposed multi-view images to enable reconstruction, segmentation, and direct object editing.

Mijin Yoo, In Cho, Subin Jeon, Jiwoo Lee and 2 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 51 on Hugging Face

Multi-Step Likelihood-Ratio Correction for Reinforcement Learning with Verifiable Rewards

NFPO augments PPO with an N-step forward trace to reduce structural bias and control variance via cumulative token likelihood ratios, improving reasoning performance.

Deokgyu Yoon, Hyungkyu Kang, Joongkyu Lee, Byeongchan Kim and 3 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Why Latent Actions Fail, and How to Prevent It

Standard reconstruction in latent action models encodes exogenous future state, but endogenous representations and action supervision prevent this failure.

Jung Min Lee, Taehyun Cho, Li Zhao, Jungwoo Lee

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Neural Quantum Spectral Operator Learning for Solving Partial Differential Equations

NVQLS proposes a hybrid quantum-classical unsupervised operator learning framework using a Legendre-Galerkin formulation to solve parametric PDEs with improved accuracy and theoretical speedups.

Chanyoung Kim, Myeonghwan Seong, Kim Yujin, Daniel Kyungdeock Park and 1 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

FLAG: Flow Policy MaxEnt-RL by Latent Augmented Guidance

FLAG augments states with flow latents to localize sampling and prevent importance weight collapse, enabling scalable high-dimensional expressive MaxEnt-RL policies with state-of-the-art performance.

Sungha Kim, Gawon Lee, Jusuk Lee, Jonghae Park and 2 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Block Sphere Vector Quantization

BlockQuant quantizes rotated vector blocks spherically to improve reconstruction and inner-product distortion over coordinate-wise methods, with unified analysis showing rotation-quantizer tradeoffs depend on distortion criteria.

Heesang Ann, Joongkyu Lee, Min-hwan Oh

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Jaguar: Fast Private CNN Inference with Power-of-Two Homomorphic Arithmetic

Jaguar uses a power-of-two ciphertext ring to replace NTT convolution with scalar accumulation and exact local truncation, cutting private CNN inference latency by up to 3.72x versus Cheetah and Rhombus.

Yewon Jeong, Nayoung Jung, Hyeri Roh, Woo-Seok Choi

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

A Locally Tokenized Generative Model for Robust Time-Series Watermarking

L-VQVAE locally tokenizes time series so each token depends on a bounded window, and LVQMark stabilizes watermark detection against post-editing attacks without sacrificing quality.

Dongbin Kim, Geonwoo Shin, Yujin Choi, Soyeon Park and 1 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

LensVLM: Selective Context Expansion for Compressed Visual Representation of Text

LensVLM lets VLMs scan compressed rendered text and selectively expand only relevant regions via learned tools, maintaining near-full accuracy at 4.3x compression and outperforming baselines up to 10.1x across text QA benchmarks.

Roy Xie, Dan Friedman, Donghan Yu, Bowen Pan and 6 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 7 on Hugging Face

Same Concept, Different Directions: Cross-Modal Feature Heterogeneity in Sparse Autoencoders

Cross-modal sparse autoencoders exhibit feature heterogeneity where shared concepts activate different latents across image and text modalities, and training modality-specific autoencoders with post-hoc alignment improves reconstruction, retrieval, and steering.

Chungpa Lee, Jihoon Kwon, Kyle Min, Jy-yong Sohn

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Distributional Alignment as a Criterion for Designing Task Vectors in In-Context Learning

Task vector design via distributional alignment with in-context learning minimizes next-token probability discrepancy, yielding a linear method that improves accuracy by 9.2% and enables cross-scale transfer.

Jihoon Kwon, Jiwon Choi, Jy-yong Sohn

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

A Composite Activation Function for Learning Stable Binary Representations

HTAF smoothly approximates Heaviside via a sigmoid-tanh composite to enable stable gradient-based training of binary neural networks, yielding interpretable ICBMs with comparable or superior accuracy.

Seokhun Park, Choeun Kim, Kwanho Lee, Sehyun Park and 2 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

MAGE: All-[MASK] Block Already Knows Where to Look in Block Diffusion LLM

MAGE uses block-diffusion's aligned all-[MASK] queries to select reusable sparse KV subsets, achieving near-lossless accuracy with up to 6.82x speedup at 128K context.

Omin Kwon, Yeonjae Kim, Doyeon Kim, Minseo Kim and 2 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Topological Periodicity Test (TopPT) via Confidence Bound of Time-Delay Embeddings

Time-delay embeddings of periodic signals are homotopy equivalent to circles, enabling TopPT, a hypothesis test with asymptotic error control for detecting periodicity via confidence-bounded persistence diagrams.

Donghyun Park, Junhyun An, Taehyoung Kim, Jisu Kim

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

OmniEgoCap: Camera-Agnostic Sequence-Level Egocentric Motion Reconstruction

OmniEgoCap uses sequence-level diffusion with geometry-aware augmentation for camera-agnostic egocentric full-body motion and shape reconstruction.

Kyungwon Cho, Jeonghyeon Na, Hanbyul Joo

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

HOPE: Hand-Object Pressure Estimation from Monocular Videos

HOPE predicts temporally evolving per-vertex hand pressure and contact from monocular video via a vertex-anchored video transformer, generalizing across objects and bare-hand settings.

Subin Jeon, Byungjun Kim, Hanbyul Joo

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

MolHIT: Advancing Molecular-Graph Generation with Hierarchical Discrete Diffusion Models

MolHIT uses hierarchical discrete diffusion and decoupled atom encoding to generate molecular graphs with near-perfect validity, surpassing 1D baselines on MOSES.

Hojung Jung, Rodrigo Hormazabal, Jaehyeong Jo, Youngrok Park and 4 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

KL for a KL: On-Policy Distillation with Control Variate Baseline

vOPD casts on-policy distillation as RL and stabilizes it via a closed-form per-token reverse KL baseline that reduces gradient variance without extra inference cost.

Minjae Oh, Sangjun Song, Gyubin Choi, Yunho Choi and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor’s Internal States

POISE predicts baselines from a model's internal states via a lightweight probe for stable, low-cost multi-domain RLVR without separate critics.

Yunho Choi, Jongwon Lim, woojin Ahn, Minjae Oh and 2 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Learning Weakly Communicating Average-Reward CMDPs: Strong Duality and Improved Regret

Strong duality holds for weakly communicating average-reward CMDPs, yielding a primal-dual algorithm with O(T^{2/3}) regret and constraint violations.

Kihyun Yu, Beomhan Baek, Dabeen Lee

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Inference-Time Vulnerability Beyond Shallow Safety: Alignment Along Generation Trajectories

Safety-aligned LLMs remain vulnerable to mid-generation token injections altering subsequent outputs, and trajectory-level alignment improves robustness beyond shallow token defenses.

Kyungmin Park, Taesup Kim

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Algorithm for Contextual Queueing Bandits with Rate-Optimal Queue Length Regret

CQB-η-2 improves contextual queueing bandit queue length regret to O~(T^-1/2) via phased exploration, matching a minimax lower bound.

Seoungbin Bae, Dabeen Lee

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Posterior Contraction Rates for sparse Kolmogorov-Arnold Networks in Anisotropic Besov Spaces

Sparse Bayesian KANs with spike-and-slab priors achieve near-minimax posterior contraction over anisotropic Besov spaces with fixed depth, adaptive smoothness, and compositional structures avoiding the curse of dimensionality.

Jeunghun Oh, Lizhen Lin, Jaeyong Lee, Kyeongwon Lee

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

EgoForce: Robust Online Egocentric Motion Reconstruction via Diffusion Forcing

EgoForce uses diffusion forcing with asymmetric noise schedules to incrementally denoise full-body motion from streaming noisy egocentric observations, outperforming online and offline methods in long-horizon reconstruction.

Inwoo Hwang, Donggeun Lim, Hojun Jang, Young Min Kim

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Distribution Matching Distillation without Fake Score Network

FSF-DMD eliminates the fake-score network in DMD for flow-map generators by using the generator's pseudo-velocity to provide reverse-divergence signals, improving ImageNet FID.

YoungJoong Kim, Deokyeong Lee, Jaesik Park

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

What Gets Measured Gets Managed: Sign-aware Recommendation Needs Sign-aware Evaluation

Sign-aware recommender systems embed valence but rank blindly, hidden by metrics that ignore disliked items; proposed signed metrics expose poor valence protection and provide trainable fixes.

Minchan Kim, Jungmin Hwang, Hyunwoo Park

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection

COMET uses multi-scale patch encoding, vector-quantized codebooks, and online adaptation to detect time-series anomalies, achieving best performance on 36 of 45 metrics.

JINWOO PARK, Hyeongwon Kang, Seung H Han, Pilsung Kang

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Toward the Goldilocks Blind Compression of Quantum States

Quantum autoencoders achieve optimal blind single-copy quantum compression with k encoder and n decoder ancillas, pinpointing the universal encoder threshold and showing isometric decoders are nearly optimal.

Hyunho Cha, Chae-Yeun Park, Jungwoo Lee

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation

DynaFLIP pre-trains dynamics-aware visual encoders via image-language-3D flow alignment, boosting robot manipulation generalization by up to 22.5%.

Jusuk Lee, Seungjae Lee, Jonghun Shin, Hoseong Jung and 5 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 9 on Hugging Face

Fast KVzip: Efficient and Accurate LLM Inference with Gated KV Eviction

Fast KVzip uses lightweight sink-attention gates to evict up to 70% of KV caches with negligible overhead, maintaining near-lossless LLM performance across long-context, code, and math tasks via forward-only, task-agnostic training.

Jang-Hyun Kim, Dongyoon Han, Sangdoo Yun

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 8 on Hugging Face · Code ★ 31

Meta$^n$: Recursive Self-Improvement through Emergent Depth

Meta^n fixes its meta-operation and recurses on growing inputs to build unbounded self-improving agent depth, outperforming prior agents across eight benchmark families including ARC-AGI-2.

Zae Myung Kim, Young-Jun Lee, Seungyeon Jwa, Dongyeop Kang

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 17 on Hugging Face

Robust Domain Generalization under Divergent Marginal and Conditional Distributions

A unified meta-learning framework minimizes decomposed risk bounds across marginal and conditional distribution shifts to achieve robust domain generalization. It achieves state-of-the-art results on standard benchmarks and challenging multi-domain long-tailed recognition settings.

Jewon Yeom, Kyubyung Chae, Hyunggyu Lim, Yoonna Oh and 2 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

From Noise to Diversity: Random Embedding Injection in LLM Reasoning

Random soft prompt injection boosts LLM math reasoning by flattening early token distributions to diversify reasoning paths and widen Pass@N without any training.

Heejun Kim, Seungpil Lee, Jewon Yeom, Jaewon Sok and 4 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Hallucination as Commitment Failure: Larger LLMs Misfire Despite Knowing the Answer

Larger instruction-tuned LLMs increasingly hallucinate despite knowing correct answers because sharpening commitment disperses probability across surface forms rather than concentrating it.

Jewon Yeom, Jaewon Sok, Heejun Kim, Seonghyeon Park and 2 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

Subject-Relative Micro-Motion and Sleep Dynamics for Near-Infrared Video Sleep Staging

ViNUSS predicts PSG-defined sleep stages directly from near-infrared video via subject-relative micro-motion and full-night dynamics, achieving 0.80 accuracy on 475 nights.

Kunmin Jang, Yourim Choi, Hun Heo, Dongik Park and 4 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

Mechanistic Circuit Identification for Controllable Data Generation

A circuit-grounded framework links mechanistic interpretability to data valuation via learnable, challenge, and alignment circuits, using them to steer generation and improve QA performance.

Nakyung Lee, Sangwoo Hong, Jungwoo Lee

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026