Author name cluster

Minsoo Kim

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

7 papers

2 author rows

ICRA Conference 2025 Conference Paper

Generating Diverse Challenging Terrains for Legged Robots Using Quality-Diversity Algorithm

Arthur Esquerre-Pourtère
Minsoo Kim
Jaeheung Park

While legged robots have achieved significant advancements in recent years, ensuring the robustness of their controllers on unstructured terrains remains challenging. It requires generating diverse and challenging unstructured terrains to test the robot and discover its vulnerabilities. This topic remains underexplored in the literature. This paper presents a Quality-Diversity framework to generate diverse and challenging terrains that uncover weaknesses in legged robot controllers. Our method, applied to both simulated bipedal and quadruped robots, produces an archive of terrains optimized to challenge the controller in different ways. Quantitative and qualitative analyses show that the generated archive effectively contains terrains that the robots struggled to traverse, presenting different failure modes. Interesting results were observed, including failure cases that were not necessarily expected. Experiments show that the generated terrains can also be used to improve RL-based controllers.

Details

NeurIPS Conference 2025 Conference Paper

InfiniPot-V: Memory-Constrained KV Cache Compression for Streaming Video Understanding

Minsoo Kim
Kyuhong Shim
Jungwook Choi
Simyung Chang

Modern multimodal large language models (MLLMs) can reason over hour-long video, yet their key–value (KV) cache grows linearly with time—quickly exceeding the fixed memory of phones, AR glasses, and edge robots. Prior compression schemes either assume the whole video and user query are available offline or must first build the full cache, so memory still scales with stream length. InfiniPot-V is the first training-free, query-agnostic framework that enforces a hard, length-independent memory cap for \textit{streaming} video understanding. During video encoding it monitors the cache and, once a user-set threshold is reached, runs a lightweight compression pass that (i) removes temporally redundant tokens via Temporal-axis Redundancy (TaR) metric and (ii) keeps semantically significant tokens via Value-Norm (VaN) ranking. Across four open-source MLLMs and four long-video and streaming-video benchmarks, InfiniPot-V cuts peak GPU memory by up to 94\%, sustains real-time generation, and matches or surpasses full-cache accuracy—even in multi-turn dialogues. By dissolving the KV cache bottleneck without retraining or query knowledge, InfiniPot-V closes the gap for on-device streaming video assistants.

PDF Details

AAAI Conference 2025 Conference Paper

RILQ: Rank-Insensitive LoRA-Based Quantization Error Compensation for Boosting 2-Bit Large Language Model Accuracy

Geonho Lee
Janghwan Lee
Sukjin Hong
Minsoo Kim
Euijai Ahn
Du-Seong Chang
Jungwook Choi

Low-rank adaptation (LoRA) has become the dominant method for parameter-efficient LLM fine-tuning, with LoRA-based quantization error compensation (LQEC) emerging as a powerful tool for recovering accuracy in compressed LLMs. However, LQEC has underperformed in sub-4-bit scenarios, with no prior investigation into understanding this limitation. We propose RILQ (Rank-Insensitive LoRA-based Quantization Error Compensation) to boost 2-bit LLM accuracy. Based on rank analysis revealing model-wise activation discrepancy loss's rank-insensitive nature, RILQ employs this loss to adjust adapters cooperatively across layers, enabling robust error compensation with low-rank adapters. Evaluations on LLaMA-2 and LLaMA-3 demonstrate RILQ's consistent improvements in 2-bit quantized inference across various state-of-the-art quantizers and enhanced accuracy in task-specific fine-tuning. RILQ maintains computational efficiency comparable to existing LoRA methods, enabling adapter-merged weight-quantized LLM inference with significantly enhanced accuracy, making it a promising approach for boosting 2-bit LLM performance.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Adversarial Environment Design via Regret-Guided Diffusion Models

Hojun Chung
Junseo Lee
Minsoo Kim
Dohyeong Kim
Songhwai Oh

Training agents that are robust to environmental changes remains a significant challenge in deep reinforcement learning (RL). Unsupervised environment design (UED) has recently emerged to address this issue by generating a set of training environments tailored to the agent's capabilities. While prior works demonstrate that UED has the potential to learn a robust policy, their performance is constrained by the capabilities of the environment generation. To this end, we propose a novel UED algorithm, adversarial environment design via regret-guided diffusion models (ADD). The proposed method guides the diffusion-based environment generator with the regret of the agent to produce environments that the agent finds challenging but conducive to further improvement. By exploiting the representation power of diffusion models, ADD can directly generate adversarial environments while maintaining the diversity of training environments, enabling the agent to effectively learn a robust policy. Our experimental results demonstrate that the proposed method successfully generates an instructive curriculum of environments, outperforming UED baselines in zero-shot generalization across novel, out-of-distribution environments.

PDF Details DOI

IROS Conference 2024 Conference Paper

RNR-Nav: A Real-World Visual Navigation System Using Renderable Neural Radiance Maps

Minsoo Kim
Obin Kwon
Howoong Jun
Songhwai Oh

We propose a novel visual localization and navigation framework for real-world environments directly integrating observed visual information into the bird-eye-view map. While the renderable neural radiance map (RNR-Map) [1] shows considerable promise in simulated settings, its deployment in real-world scenarios poses undiscovered challenges. RNR-Map utilizes projections of multiple vectors into a single latent code, resulting in information loss under suboptimal conditions. To address such issues, our enhanced RNR-Map for real-world robots, RNR-Map++, incorporates strategies to mitigate information loss, such as a weighted map and positional encoding. For robust real-time localization, we integrate a particle filter into the correlation-based localization framework using RNR-Map++ without a rendering procedure. Consequently, we establish a real-world robot system for visual navigation utilizing RNR-Map++, which we call "RNR-Nav. " Experimental results demonstrate that the proposed methods significantly enhance rendering quality and localization robustness compared to previous approaches. In real-world navigation tasks, RNR-Nav achieves a success rate of 84. 4%, marking a 68. 8% enhancement over the methods of the original RNR-Map paper.

Details

IROS Conference 2023 Conference Paper

Real-Time Motion Planning Framework for Autonomous Vehicles with Learned Committed Trajectory Distribution

Minsoo Kim
Seho Shin
Joonwoo Ahn
Jaeheung Park

This study proposes a realtime motion planning framework that leverages the prediction of a portion of the optimal trajectory for sampling-based anytime planning algorithms. Existing algorithms predict the entire optimal path and bias random samples toward it for fast path planning. However, these algorithms may not be suitable for realtime frameworks because the bias-sampling strategy should consider the sequential nature of realtime execution. Therefore, the proposed algorithm predicts a portion of the optimal path, known as the committed trajectory, step by step as a probability distribution using a neural network. This distribution is then used in a sampling-based anytime planning algorithm as a non-stationary way of biasing random samples. The proposed algorithm can sequentially plan the near-optimal motion, al-lowing the vehicle to reach the desired goal pose in a timely and accurate manner. In various test parking scenarios, the proposed algorithm reduces the parking time by approximately 38% compared with conventional motion planning algorithms and by 10% compared with another realtime framework that biases samples toward the entire optimal trajectory.

Details

NeurIPS Conference 2023 Conference Paper

Token-Scaled Logit Distillation for Ternary Weight Generative Language Models

Minsoo Kim
Sihwa Lee
Janghwan Lee
Sukjin Hong
Du-Seong Chang
Wonyong Sung
Jungwook Choi

Generative Language Models (GLMs) have shown impressive performance in tasks such as text generation, understanding, and reasoning. However, the large model size poses challenges for practical deployment. To solve this problem, Quantization-Aware Training (QAT) has become increasingly popular. However, current QAT methods for generative models have resulted in a noticeable loss of accuracy. To counteract this issue, we propose a novel knowledge distillation method specifically designed for GLMs. Our method, called token-scaled logit distillation, prevents overfitting and provides superior learning from the teacher model and ground truth. This research marks the first evaluation of ternary weight quantization-aware training of large-scale GLMs with less than 1. 0 degradation in perplexity and achieves enhanced accuracy in tasks like common-sense QA and arithmetic reasoning as well as natural language understanding. Our code is available at https: //github. com/aiha-lab/TSLD.

PDF Details