Author

L.J.

@zljdanceholic

articles: 88
total claps: 81
top claps: 67
avg read: 2.5 min
active: 2025-10-17 → 2026-07-20

Wiki topics

LLM · Large Language Models · 54 AI · AI · General · 36 EDU · Education & Learning · 29 MM · Multimodal & Generative Media · 16 ML · Machine Learning · 11 FT · Fine-tuning & Adaptation · 10 AGT · AI Agents · 9 CRM · Email & CRM · 8

Top tags

#llm ·53 #ai ·36 #2026 ·31 #arxiv ·20 #iclr ·19 #aaai ·17 #miccai ·13 #medical-ai ·8 #neurips ·7 #oral-paper ·7

SOC AI SCI

LJ L.J.

Math Proves It: Transformer Heads Can Either Know “Where” or “What” — But Never Both

We all know that multi-head attention in Transformers involves a division of labor, where different heads look at different things. But…

May 30, 2026 · 2 min read · 67 claps #iclr #2026 #llm

Math Proves It: Transformer Heads Can Either Know “Where” or “What” — But Never Both

SOC HUM MDA AI

LJ L.J.

Can NeurIPS 25 oral RLVR really improve reasoning ability?

Sharing a NeurIPS 2025 Oral Paper. A Collaboration between Tsinghua University and Shanghai Jiao Tong University

Dec 18, 2025 · 2 min read · 5 claps #neurips #reinforcement-learning #rlvr

Can NeurIPS 25 oral RLVR really improve reasoning ability?

MED HUM SOC

LJ L.J.

The Right to be Forgotten in Medical AI: How Maverick Solves the Federated Unlearning Crisis

In the world of medical Federated Learning (FL), the “Right to be Forgotten” isn’t just a legal checkbox — it’s a technical nightmare…

Feb 09, 2026 · 3 min read · 4 claps #miccai #oral-paper #federated-learning

The Right to be Forgotten in Medical AI: How Maverick Solves the Federated Unlearning Crisis

AI TCH HUM

LJ L.J.

Different Models, Same Logic: How LLMs Converge on Shared Reasoning Patterns

A team from Shanghai Jiao Tong University just dropped a new paper on arXiv asking a question that’s been quietly bugging the field: when…

Jun 13, 2026 · 2 min read · 2 claps #arxiv #llm #ai

Different Models, Same Logic: How LLMs Converge on Shared Reasoning Patterns

AI SCI

LJ L.J.

The $1 Review: What Happens When AI Audits 23,000 AAAI Submissions?

The peer review system in academia has been redlining for years. To put the scale into perspective: AAAI 2026 received over 20,000…

May 13, 2026 · 2 min read · 2 claps #aaai #2026 #arxiv

The $1 Review: What Happens When AI Audits 23,000 AAAI Submissions?

TCH AI

LJ L.J.

Beyond the Reversal Curse: How Diffusion LLMs Secretly Solve the Long-Context Problem

This AAAI 2026 paper presents the first systematic study of long-context capabilities in Diffusion Large Language Models (dLLMs)…

Apr 27, 2026 · 2 min read · 1 claps #aaai #2026 #llm

Beyond the Reversal Curse: How Diffusion LLMs Secretly Solve the Long-Context Problem

ECO ART AI

LJ L.J.

Beyond Pairwise: How Sapienza’s GRAM is Redefining Multimodal Alignment at ICLR 2025

In the rapidly evolving world of multimodal AI, the “CLIP-style” approach has long been the gold standard. We use cosine similarity to pull…

Mar 22, 2026 · 3 min read · 0 claps #iclr #gram #gramian

Beyond Pairwise: How Sapienza’s GRAM is Redefining Multimodal Alignment at ICLR 2025

TCH AI SOC

LJ L.J.

Knowing Which Skills Fine-Tuning Will Break — Before You Fine-Tune

There’s a frustrating law that governs almost all post-training: you push one capability up, and something else quietly slides down. You…

Jun 07, 2026 · 3 min read · 0 claps #iclr #2026 #sft

Knowing Which Skills Fine-Tuning Will Break — Before You Fine-Tune

HUM MKT MED AI

LJ L.J.

Solving the “Missing Modality” Headache in Brain Tumor Segmentation: A Deep Dive into DC-Seg

For anyone working in medical imaging, the “missing modality” problem isn’t just a theoretical edge case — it’s the reality. Whether…

Feb 02, 2026 · 3 min read · 0 claps #miccai #segmentation #contrastive-learning

Solving the “Missing Modality” Headache in Brain Tumor Segmentation: A Deep Dive into DC-Seg

AI HUM

LJ L.J.

Why a Language Model Gives Different Moral Answers in Different Languages — and Why “Culture” Isn’t…

There’s a single-author paper out of Nanyang Technological University with a finding that sounds simple but gets stranger the longer you…

Jun 06, 2026 · 3 min read · 0 claps #arxiv #2026 #llm

Why a Language Model Gives Different Moral Answers in Different Languages — and Why “Culture” Isn’t…

HUM AI

LJ L.J.

Beyond Blind Steps: Rethinking Learning Rate Warm-Up with SAWU

In the world of fine-tuning Large Language Models (LLMs), linear warm-up is the industry’s “default setting.” We start with a tiny learning…

Apr 06, 2026 · 3 min read · 0 claps #aaai #2026 #warm-up

Beyond Blind Steps: Rethinking Learning Rate Warm-Up with SAWU

AI HUM

LJ L.J.

Bridging the Reality Gap in Autonomous Driving: Why Active Learning is the Secret Sauce for LiDAR…

LiDAR sensors in autonomous vehicles generate hundreds of thousands of points per second, creating a sparse, flickering 3D representation…

Apr 05, 2026 · 3 min read · 0 claps #aaai #2026 #autonomous-driving

Bridging the Reality Gap in Autonomous Driving: Why Active Learning is the Secret Sauce for LiDAR…

LJ L.J.

From NeurIPS Rejection to ICLR 2025 Oral: Why KAN is More Than Just a Hype

The “Yes we KAN” craze that took over the AI world in the first half of 2024 had a bit of a rollercoaster ride. After reportedly getting a…

Mar 14, 2026 · 3 min read · 0 claps #iclr #kan #neurips

From NeurIPS Rejection to ICLR 2025 Oral: Why KAN is More Than Just a Hype

LJ L.J.

Solving the Long-Horizon Jitter: Inside LatentVLA, the New Standard for Bimanual Robot Manipulation

The world of Vision-Language-Action (VLA) models has been grappling with a persistent headache: long-horizon tasks. While moving a cup from…

Apr 28, 2026 · 3 min read · 0 claps #aaai #vla #diffusion-models

Solving the Long-Horizon Jitter: Inside LatentVLA, the New Standard for Bimanual Robot Manipulation

HUM MKT ART

LJ L.J.

The Boundary Problem: Why Generative Priors are the New Frontier in Medical Image Segmentation

In the world of medical image segmentation, the real battle isn’t just identifying an organ — it’s drawing the line. Literally.

Feb 08, 2026 · 3 min read · 0 claps #miccai #oral-paper #segmentation

The Boundary Problem: Why Generative Priors are the New Frontier in Medical Image Segmentation

MKT

LJ L.J.

SAM2 Meets Dentistry: Bridging the 2D-to-3D Gap in Automated Tooth Segmentation

3D tooth segmentation is a classic yet notoriously stubborn challenge in digital dentistry. While the goal is simple — identifying and…

Apr 04, 2026 · 3 min read · 0 claps #sam2 #aaai #segmentation

SAM2 Meets Dentistry: Bridging the 2D-to-3D Gap in Automated Tooth Segmentation

AI TCH HUM

LJ L.J.

The Ghost in the Machine: Revisiting a Decade-Old Warning on Social Bots

I recently revisited a seminal review on social bots published in Communications of the ACM nearly a decade ago. It’s a striking read —…

Feb 12, 2026 · 3 min read · 0 claps #cacm #machine-learning #bots

The Ghost in the Machine: Revisiting a Decade-Old Warning on Social Bots

AI MDA

LJ L.J.

The New Frontier of LLMs: Why “In-Place TTT” is the Breakthrough We’ve Been Waiting For

Modern Large Language Models (LLMs) generally follow a rigid, two-phase lifecycle: first, we compress vast amounts of human knowledge into…

Apr 09, 2026 · 3 min read · 0 claps #llm #ttt #test-time-training

The New Frontier of LLMs: Why “In-Place TTT” is the Breakthrough We’ve Been Waiting For

AI TCH

LJ L.J.

Bridging the Gap: Why We Should Stop Choosing Between Model Cascading and Speculative Decoding

In the world of LLM inference optimization, two heavyweights have long dominated the conversation: Model Cascading and Speculative…

Mar 13, 2026 · 3 min read · 0 claps #iclr #oral-paper #ai

Bridging the Gap: Why We Should Stop Choosing Between Model Cascading and Speculative Decoding

LJ L.J.

Diffusion’s Hidden Dimension: Why “What We Generate First” Changes Everything

In a recent paper from Fei-Fei Li’s team, the core conversation shifts toward a fundamental but often overlooked question in diffusion…

Feb 15, 2026 · 3 min read · 0 claps #arxiv #feifei-li #stable-diffusion

Diffusion’s Hidden Dimension: Why “What We Generate First” Changes Everything

AI SCI BIO

LJ L.J.

Do LLMs Have Functional Brain Regions? Mapping the Neural “Geography” of Logic and Language

When a Large Language Model (LLM) solves a complex calculus problem or debugs a Python script, is it using the same pool of “gray matter,”…

May 08, 2026 · 3 min read · 0 claps #aaai #2026 #llm

Do LLMs Have Functional Brain Regions? Mapping the Neural “Geography” of Logic and Language

LJ L.J.

Bridging the “Spatial Blindness” Gap: How VEGA-3D Leverages Video Generation for 3D Understanding

Modern Multi-modal Large Language Models (MLLMs) are surprisingly adept at “chatting” with images, yet they often hit a wall when faced…

Mar 23, 2026 · 3 min read · 0 claps #arxiv #3d-understanding #llm

Bridging the “Spatial Blindness” Gap: How VEGA-3D Leverages Video Generation for 3D Understanding

LJ L.J.

AI-assisted Paper Searching: arXivSub

🔥 arXivSub is officially online! It’s been almost three months from the end of July to now. Thank you to everyone who participated in the…

Oct 17, 2025 · 3 min read · 0 claps #arxiv #research #ai

AI MED SOC

LJ L.J.

WSI-Agents: Can Multi-Agent Collaboration Bridge the Performance Gap in Digital Pathology?

In the world of medical AI, we often face a frustrating trade-off: Generalization vs. Specialization. While Generalist Multimodal Large…

Feb 16, 2026 · 3 min read · 0 claps #miccai #agents #medical-ai

WSI-Agents: Can Multi-Agent Collaboration Bridge the Performance Gap in Digital Pathology?

SOC HUM AI

LJ L.J.

NeurIPS25 oral DiT training speed up new SOTA: REG

A recent collaboration between Nankai University and Harvard University has produced a standout NeurIPS 2025 Oral paper that might just…

Jan 12, 2026 · 3 min read · 0 claps #neurips #neurips-oral #dit