Hsin-Ling Hsu

Undergrad junior @ NCCU. Trustworthy VLMs/LLMs; AI for Healthcare.

[Andrew Conner @connerdelights]

Terence Tao posted his ChatGPT session trying to understand the Jacobian conjecture counterexample. It's so lovely reading a slice of how his mind works, the connections he's making, etc.

https://t.co/wu6CcAk3V6

[Sam Altman @sama]

we had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this.

https://t.co/2o2VfR6PIa

[Hsin-Ling Hsu @JustinHsu99]

[COLM 2026] Thrilled to receive this surprise right before the end of my undergrad junior year: our paper MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs has been accepted to COLM 2026! 🎉🎉

This is my first first-author paper at a top-tier conference main track (I had an ACL paper last year, but industry track, though ACL industry was around a 25% acceptance rate too, so still pretty competitive xD). I feel incredibly lucky to have produced this result together with the professors, physicians, and collaborators/seniors at the University of Michigan and Far Eastern Memorial Hospital. Heartfelt thanks to everyone on the team. With COLM's acceptance rate at 29% this year and submissions surging, it was even more competitive than last year, which makes getting in all the more exciting.

See you all in San Francisco this October! 🌉

---

Paper overview:

Most medical LLMs are evaluated in a static, single-turn setting: give the model a complete record and have it predict the disease/ICD directly. But real, complex clinical settings aren't always like that. A physician starts from the chief complaint, then step by step orders tests, interprets results, updates the differential diagnosis, and commits to a final diagnosis once confident.

❓ When you turn diagnosis into a truly multi-turn, active process, how do current LLMs do? Even SOTA models run into three major problems: ungrounded test ordering, unreliable update, and degraded coherence. Existing data mostly teaches models to reason when information is complete, but not how to act when the evidence keeps changing.
❓ So how do we close this gap? We propose MedAction, which has LLMs interact with a simulated clinical environment to generate multi-turn diagnostic trajectories, then filters trajectory quality using two newly proposed KG metrics (DTC and RAC). The 8B model trained on it beats a 235B teacher model, reaches SOTA among open-source models, and earned recognition from clinical physicians.

Full paper: https://t.co/X2lH26l8wD

#COLM2026 #LLM #MedicalAI #ClinicalReasoning #AIforHealthcare #MachineLearning

[Jun-En Ding @CYh4lhDXrLDakuk]

🎉[COLM'26] Excited to share that our paper “MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs” has been accepted to COML'26!
We introduce MedAction-32K to train LLMs for active diagnosis through iterative test ordering and evidence-based diagnostic updates.#COLM2026

[Hsin-Ling Hsu @JustinHsu99]

[COLM 2026] Thrilled to receive this surprise right before the end of my undergrad junior year: our paper MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs has been accepted to COLM 2026! 🎉🎉

This is my first first-author paper at a top-tier conference main track (I had an ACL paper last year, but industry track, though ACL industry was around a 25% acceptance rate too, so still pretty competitive xD). I feel incredibly lucky to have produced this result together with the professors, physicians, and collaborators/seniors at the University of Michigan and Far Eastern Memorial Hospital. Heartfelt thanks to everyone on the team. With COLM's acceptance rate at 29% this year and submissions surging, it was even more competitive than last year, which makes getting in all the more exciting.

See you all in San Francisco this October! 🌉

---

Paper overview:

Most medical LLMs are evaluated in a static, single-turn setting: give the model a complete record and have it predict the disease/ICD directly. But real, complex clinical settings aren't always like that. A physician starts from the chief complaint, then step by step orders tests, interprets results, updates the differential diagnosis, and commits to a final diagnosis once confident.

❓ When you turn diagnosis into a truly multi-turn, active process, how do current LLMs do? Even SOTA models run into three major problems: ungrounded test ordering, unreliable update, and degraded coherence. Existing data mostly teaches models to reason when information is complete, but not how to act when the evidence keeps changing.
❓ So how do we close this gap? We propose MedAction, which has LLMs interact with a simulated clinical environment to generate multi-turn diagnostic trajectories, then filters trajectory quality using two newly proposed KG metrics (DTC and RAC). The 8B model trained on it beats a 235B teacher model, reaches SOTA among open-source models, and earned recognition from clinical physicians.

Full paper: https://t.co/X2lH26l8wD

#COLM2026 #LLM #MedicalAI #ClinicalReasoning #AIforHealthcare #MachineLearning

[Tanat Tonguthaisri @gastronomy]

OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization: Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording. W… deployments.https://t.co/2l3VeYpXwt

[Cryptography and Security Papers @FSFG]

OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization

Jerry Wang, Hsin-Ling Hsu, Yi-Cheng Lai, Nai-Chia Chen, Fang Yu
https://t.co/IiP2Ntm4Ce [𝚌𝚜.𝙲𝚁 𝚌𝚜.𝙲𝙻]

[Muhan Gao @muhan_gao]

🤖 We often talk about “context rot”: LLMs get worse as context grows.

But once distracting information enters, is it just “a bit more noise → a bit worse performance”?

Our #ICML2026 paper finds: no! 🤯 Instead, we reveal a striking "First Drop of Ink" effect: the first very few hard distractors do almost all of the damage, exactly like how one drop of ink clouding clear water.

Paper link: https://t.co/s3zg49imUr

[Conference on Language Modeling @COLM_conf]

The discussion period for COLM 2026 is underway! We're sharing a CDF of average review scores. Note that final decisions will reflect deliberation by ACs and PCs, so these are only meant to be a heuristic guideline to give you a sense of where your papers stand. Good luck!

[Natural Language Processing Papers @HEI]

MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs

Hsin-Ling Hsu, Zizheng Wang, Donghua Zhang, Nai-Chia Chen, Jerry Wang, Jun-En Ding, Chia-Hsuan Hsu, Guoan Wang, Feng Liu, Fang-Ming Hung, Chenwei Wu, Liyue Shen
https://t.co/4sBYSemwwu [𝚌𝚜.𝙲𝙻 𝚌𝚜.𝙰𝙸]

[Machine Learning (ML) Papers @Memoirs]

WARP: Guaranteed Inner-Layer Repair of NLP Transformers

Hsin-Ling Hsu, Min-Yu Chen, Nai-Chia Chen, Yan-Ru Chen, Yi-Ling Chang, Fang Yu
https://t.co/e4eoJ8tYfo [𝚌𝚜.𝙻𝙶 𝚌𝚜.𝙰𝙸]

[Sharon Li @SharonYixuanLi]

When evaluating LVLMs, should we really be asking:
“Did the model get the right answer?”
or rather
“Did the model truly integrate the visual input?”

LVLMs can rely on shortcuts learned from the underlying language model, aka language prior.
In our #ICLR2026 paper, we attempt to understand this phenomenon at a deeper, representation-level.
📄 “Understanding Language Prior of LVLMs by Contrasting Chain-of-Embedding”.
https://t.co/J8NgQ7AyGc

-------
1/ Problem: LVLMs often ignore visual evidence

While LVLMs perform well on many benchmarks, they sometimes rely on language patterns rather than actual images.
A simple example: show a model a green banana, and it may confidently describe it as “ripe and yellow” ---because that’s the most common linguistic pattern it has learned. 🍌

This raises a central question:
Where inside the model does visual information begin to influence its reasoning?

2/ Motivation: Output-level probes fall short

Most analyses inspect outputs, e.g., by removing the image or comparing predictions.
But these methods cannot reveal when the model starts integrating vision and how strongly visual signals affect internal states.

To address this, we need a representation-driven perspective. 🔍

3/ Approach: Contrasting Chain-of-Embedding (CoE)
We trace hidden representations across the model’s depth for the same prompt:

•once with the image
•once without the image

By comparing these trajectories layer by layer, we identify the exact point where visual input begins shaping the model’s internal computation.

This leads to the discovery of the Visual Integration Point (VIP) ✨--- the layer at which the model “starts seeing.”

We then define Total Visual Integration (TVI), a metric that quantifies how much visual influence accumulates after the VIP.

4/ Findings across 10 LVLMs and 6 benchmarks

Across 60 evaluation settings, we observe:
• VIP consistently appears across diverse architectures
• Pre-VIP → representations behave like a language-only model
• Post-VIP → visual signals increasingly reshape the embedding pathway
• TVI correlates strongly with actual visual reasoning performance
• TVI outperforms attention- and output-based proxies at identifying language prior

TVI thus offers a more principled indicator of whether a model actually uses the image.

5/ Impact: A new lens on multimodal behavior

Our framework has a few practical benefits. It enables (1) diagnosing over-reliance on language prior, (2) comparing LVLM architectures more rigorously, (3) informing better training and alignment strategies, and (4) improving robustness and grounding in real-world tasks.

Shout out to my students for this insightful work: Lin Long, @Changdae_Oh, @seongheon_96 🌻

Please check out our paper for more details!

[Shangbin Feng @shangbinfeng]

Introducing MoCo, the first open-source Python package for model collaboration, where multiple LLMs collaborate, compose, and complement each other.

Code: https://t.co/qjW5ZCZ0s9
Paper: https://t.co/yn6ul2hHeY

[Akari Asai @AkariAsai]

Thrilled to share: OpenScholar - our work on scientific deep research agents for reliable literature synthesis -has been accepted to Nature! 🎉 Huge thanks to collaborators across institutions who made this possible!

[OpenAI @OpenAI]

Introducing Prism, a free workspace for scientists to write and collaborate on research, powered by GPT-5.2.

Available today to anyone with a ChatGPT personal account: https://t.co/9mTLAbxPdH

[Kuan-Hao Huang @kuanhaoh_]

We're organizing the 1st Texas NLP Symposium! A one-day workshop bringing together NLP researchers across Texas and beyond to share ongoing work.

📅 April 3, 2026
📍 Texas A&M University

📣 Call for Papers
⏰ Deadline: Feb 20
🔗 https://t.co/fZCFz2dsOt

Consider submitting!

[OpenAI @OpenAI]

Introducing ChatGPT Health — a dedicated space for health conversations in ChatGPT. You can securely connect medical records and wellness apps so responses are grounded in your own health information.

Designed to help you navigate medical care, not replace it.

Join the waitlist to get early access.

https://t.co/MdpqDg7Ecg

[NTCIR @NTCIR]

== CALL FOR NTCIR-19 TASK PARTICIPATION ==
#NTCIR19 now calls for task participation.
For the details, please visit https://t.co/5Sh0qoI9SC
Please join us for #NTCIR!

[Conference on Language Modeling @COLM_conf]

COLM 2026 is just around the corner! Mark your calendars for:

💡Abstract deadline: Thursday, March 26, 2026
📄Full paper submission deadline: Tuesday, March 31, 2026

Call for papers in thread (website coming soon).

[Conference on Language Modeling @COLM_conf]

COLM is going to San Francisco for 2026!

🗓️Dates: October 6-9, 2026
🏨Venue: Hilton San Francisco Union Square

Website and CFPs for papers and workshops coming up soon!