CollabVR: Collaborative Video Reasoning with Vision-Language and Video Generation Models

CollabVR pairs vision-language models with video generation models in closed-loop step-level planning and verification, reducing drift and simulation errors for major video reasoning gains.

Joowon Kim, Seungho Shin, Joonhyung Park, Eunho Yang

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 71 on Hugging Face · Code ★ 10

Attention Sinks as Spectral Spikes: A Mechanism Analysis of Gated Attention

Seojin Kim, Yehjin Shin, Noseong Park

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

HAPACT: A Benchmark For Human-Centric Physical Impact Localization in Movies

Youngrae Kim, Yejin Jang, Eunho Kim, Youjin Sung and 3 more

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

Towards Error-Free EHRs: Reasoning-Intensive Consistency Verification Between Clinical Notes and Structured Tables in Electronic Health Records

EHR-ReasonCon introduces a reasoning-intensive benchmark for clinical note-table consistency verification, and EHR-Inspector achieves state-of-the-art results via LLM-based verification with table exploration.

Yeonsu Kwon, Jiho Kim, Junseong Choi, Paloma Rabaey and 9 more

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

AtomMOF: All-Atom Flow Matching for MOF-Adsorbate Structure Prediction

AtomMOF uses all-atom flow matching to predict MOF and adsorbate structures directly from 2D graphs, improving match rates and sampling efficiency.

Nayoung Kim, Honghui Kim, Sihyun Yu, Minkyu Kim and 2 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · Code ★ 19

Atom-level Protein Representation Learning Improves Protein Structure Prediction

TriProRep pretrains structure-aware protein representations via joint amino-acid, backbone, and full-atom token recovery, improving homodimer co-folding, interaction prediction, and monomer structure prediction over sequence-only models.

Taewon Kim, Hyosoon Jang, Hyunjin Seo, Seonghwan Seo and 5 more

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

EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries

EHRNote-ChatQA introduces multi-turn clinical QA over longitudinal discharge summaries, showing LLMs struggle with evidence grounding and multi-turn error accumulation.

Jiyoun Kim, Muhan Yeo, Eunhye Jang, Jeewon Yang and 13 more

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

Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning

ToMA uses persistent homology to align cross-modal manifold edges via image-text pairs, improving semi-supervised vision-language learning in specialized domains.

Junwon You, Mihyun Jang, Sangwoo Mo, Jae-Hun Jung

Sydney Poster Session 5, Thu, Dec 10, 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

ECG-Reasoning-Benchmark: A Benchmark for Evaluating Clinical Reasoning Capabilities in ECG Interpretation

A new benchmark reveals current multimodal AI fails at multi-step ECG reasoning, achieving near-zero completion in linking clinical criteria to visual signal evidence.

Jungwoo Oh, Hyunseung Chung, Junhee Lee, Min-Gyu Kim and 5 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 6 on Hugging Face · Code ★ 18

A Systematic Evaluation of Co-folding Model Representations for Small-Molecule Learning

Boltz2's protein-ligand co-folding representations match or outperform standalone models on ADMET, generative modeling, and ligand optimization tasks while complementing conventional molecular supervision.

Hyosoon Jang, Hyunjin Seo, Honghui Kim, Taewon Kim and 3 more

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