sorryhyun

Knowledge gains its value when shared.

2023202420252026AI TrendArxivCOLMConference ReviewCritiqueFoundationalGeneralizationICLRICMLIndustryLLMLong ContextMiscNeurIPSOpen QuestionTheory
2026-01-01

NeurIPS 2025: Selected Papers

Selected NeurIPS 2025 papers across novel architectures, latent analysis, challenges to traditional beliefs, diffusion/SSM, and simple methods.

Conference ReviewNeurIPS2025
2026-01-01

Other Conferences (2026)

Reviews from ICLR 2026 selected papers — Why low-precision transformer training fails (Flash Attention analysis), and more.

Conference ReviewICLR2026
2025-12-31

Notable Arxiv Papers (2025)

Standout 2025 arxiv preprints — Predictable Scale (optimal hyperparameter scaling laws for LLM pretraining), and more.

Arxiv2025
2025-12-31

Cute Figures

A collection of memorable figures from papers — honesty/helpfulness in LLMs, training-loss zero, mechanistic explanations, and more.

Misc
2025-12-31

Research from Industry

Notable papers from industry labs — DeepSeek-OCR (contexts as optical compression) and more.

IndustryLLM
2025-10-01

Other Conferences (2025)

Reviews from ICLR 2025 orals — Unlearning-based Neural Interpretations and more.

Conference ReviewICLR2025
2025-07-01

ICML 2025: Oral & Spotlighted

ICML 2025 oral and spotlighted papers, organized by theme — new interpretations, consolidated metrics, observation-driven approaches.

Conference ReviewICML2025
2025-06-01

Generalization via Memorization

Reviewing the counter-intuitive relationship between memorization and generalization in deep networks — from rethinking-generalization (ICLR 2017) onward.

Open QuestionTheoryGeneralization
2025-03-01

Long-Range Context for LLMs

Survey of architectures and training strategies that extend LLM context — efficient transformers, SSMs, and benchmarks.

AI TrendLLMLong Context2025
2025-01-01

Disputed Trends

Trends in deep learning that may be artifacts of evaluation choices — emergent abilities and how metric design can manufacture or erase them.

AI TrendCritiqueTheory
2025-01-01

NeurIPS 2024: Oral Session

NeurIPS 2024 oral session papers grouped by theme — rethinking motivations, interpretability for LLMs, novel architectures, applications.

Conference ReviewNeurIPS2024
2024-12-31

Notable Arxiv Papers (~2024)

Selected arxiv preprints worth reading from 2024 and earlier.

Arxiv2024
2024-10-01

Other Conferences (2024)

Reviews from COLM 2024 and ICLR 2024 outstanding papers — TOFU unlearning, fair long-sequence comparisons, and more.

Conference ReviewICLRCOLM2024
2024-07-01

ICML 2024: Oral & Spotlighted

Reviews of ICML 2024 oral and spotlighted papers, with trend context and motivations laid out before each paper.

Conference ReviewICML2024
2024-06-01

Differential Equations for LLMs

From Neural ODE to differential-equation-style formulations of LLM training and inference.

AI TrendLLMTheory2024
2024-01-01

Test-of-Time Candidates

Older papers worth revisiting — both for industry impact and for the problem-solving strategies behind them. Starting with the Neural Tangent Kernel.

FoundationalTheory
2024-01-01

Miscellaneous Notes

Short notes on papers that don't fit elsewhere — starting with Similarity of Neural Network Representations Revisited.

MiscTheory
2024-01-01

NeurIPS 2023: Oral & Spotlight (LLM)

Reviews of NeurIPS 2023 oral and spotlight papers focused on large language models.

Conference ReviewNeurIPSLLM2023