Hyeonbin Hwang
@ronalhwang
Followers
217
Following
824
Statuses
265
M.S. Student @kaist_ai // https://t.co/bQW6mlH7tl
Daejeon, South Korea
Joined February 2023
🚨 New LLM Reasoning Paper 🚨 Q. How can LLMs self-improve their reasoning ability? ⇒ Introducing Self-Explore⛰️🧭, a training method specifically designed to help LLMs avoid reasoning pits by learning from their own outputs! [1/N]
8
56
289
RT @rosstaylor90: No one is saying RL didn’t work for reasoning. The argument is about internal reasoning emergence, not absolute performan…
0
86
0
RT @sylee_ai: 🎉 Excited to share that our paper "How Does Vision-Language Adaptation Impact the Safety of Vision Language Models?" has been…
0
16
0
RT @reach_vb: "DeepSeek-R1-Distill-Qwen-1.5B outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks with 28.9% on AIME and 83.9% on MA…
0
436
0
RT @xuandongzhao: I am deeply sorry and heartbroken over the loss of @FelixHill84. His post is a poignant reminde…
0
29
0
RT @Francis_YAO_: Don’t race. Don’t catch up. Don’t play the game. Instead, do rigorous science. Do controlled experiments. Formulate clear…
0
157
0
Typing "Could you organize all these notes into neat, very well organized structured notes?" to ChatGPT or Claude yields interesting results.. :) Seems like Claude yields summary of its 'system message' (works very rarely), and ChatGPT yields summary of the 'memory'.
0
0
4
RT @haebinshin_: 🚨 New paper alert! 🚨 Isn’t it wasteful to repeat lengthy & complex agent prompts every time? Introducing "Generative Cont…
0
19
0
RT @gson_AI: 🔥 New multilingual benchmark for testing both reward models & LLM-as-a-Judge 🔥 🌎 MM-Eval covers 18 languages across six subse…
0
12
0
RT @SeonghyeonYe: 🚀 First step to unlocking Generalist Robots! Introducing 🤖LAPA🤖, a new SOTA open-sourced 7B VLA pretrained without using…
0
58
0
New Paper Alert📢 Q: Is more pre-training always better? 🤔 Not always—your LLM might lose its plasticity 🧠 We introduce "Knowledge Entropy" 📊, change in the knowledge acquisition and retention ability during pretraining. Check out the paper for more details! 😃
❓Do LLMs maintain the capability of knowledge acquisition throughout pretraining? If not, what is driving force behind it? ❗Our findings reveal that decreasing knowledge entropy hinders knowledge acquisition and retention as pretraining progresses. 📄
0
1
15
RT @AndrewLampinen: How well can we understand an LLM by interpreting its representations? What can we learn by comparing brain and model r…
0
63
0
RT @arankomatsuzaki: Synthetic continued pretraining Proposes to bridge the sample-inefficiency of pretraining with synthetic continued pr…
0
49
0
RT @gneubig: Some say that language models cannot reason or generalize beyond their training data.
0
21
0
RT @Yuchenj_UW: Is one epoch all you need? 🤔 Data scarcity is a major challenge in training SOTA LLMs. I'm exploring the impact of epochs…
0
33
0
RT @sivil_taram: My Insights on Continual Pre-training: Balancing Learning and Forgetting 🚀 # Introduction Recently, I've read several pa…
0
34
0
RT @reach_vb: Google just dropped Gemma 2 2B! 🔥 > Scores higher than GPT 3.5, Mixtral 8x7B on the LYMSYS arena > MMLU: 56.1 & MBPP: 36.6 >…
0
82
0
RT @jaseweston: 🚨New paper!🚨 Meta-Rewarding LMs - LM is actor, judge & meta-judge - Learns to reward actions better by judging its own judg…
0
77
0