arXiv subjects

Minseon Kim

Publications and source records attributed to Minseon Kim.

At least 19 recordsLinked to original sources

arXiv2026

Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety Routing

The prevailing paradigm in large language model (LLM) alignment operates via erasure, filtering unsafe data or training models to strictly refuse harmful prompts. While effective at reducing immediate toxicity, this approach fundamentally constricts the model's epistemological scope, resulting in over-cautious systems that output uninformative blanket refusals to sensitive yet benign queries. In this work, we challenge the orthodoxy that unsafe data must be discarded. We propose a dialectical approach to alignment, positing that unsafe data encodes rich, domain-specific knowledge critical for nuanced, safe, and informative generation. To operationalize this, we introduce SafeMoE, a Mixture-of-Experts (MoE) framework that isolates unsafe knowledge into domain-specific Low-Rank Adapters (LoRA experts) trained exclusively on harmful corpora. To synthesize safety from these unsafe primitives, we train a lightweight gating network using a minimal, highly curated set of safe-informative responses (as few as 100 samples per domain). During inference, this router dynamically orchestrates the unsafe experts, effectively steering the generation trajectory toward safe outcomes while harnessing the experts' deep domain knowledge. Extensive empirical evaluations across stringent safety benchmarks demonstrate that SafeMoE is not only safer, achieving over a 20% relative improvement in safe response rate over the strongest baseline (RealSafe-R1 on PKU-SafeRLHF; more than a 15% absolute gain), but also produces more informative responses when safety and harmfulness are of paramount concern. Furthermore, the routing mechanism exhibits strong zero-shot generalization to unseen domains and broader safety tasks without domain-specific supervision. Our findings suggest a paradigm shift in alignment: true safety requires not the masking of unsafe knowledge, but its controlled integration.

cs.LG↗

arXiv2026

KV-streams for Efficient Compaction in Agentic Reinforcement Learning

Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most popular mechanism to alleviate this issue, keeping GPU memory constant for a given trace. Unfortunately, most compaction strategies rely on prefilling the LLM context many times over, hindering training throughput. To alleviate this bottleneck and enable efficient trainable compaction, we propose KV-streams, a plug-and-play strategy compatible with any compaction strategy that substantially increases throughput while showing no evidence of hindering performance. KV-streams enable scalable compaction by streaming the KV cache forward rather than flushing it after each compaction. We show that KV-streams enable three different compaction strategies, achieving a 2.6 to 5x wall-clock speedup in training. Beyond efficiency, we find that the streamed KV cache can act as a recurrent state, carrying forward information that has long since disappeared from the context. Specifically, in a controlled setting we show that, contrary to prior work, RL alone is all that is needed for this behavior to emerge. Overall, we show KV-streams to be an efficient and lightweight plug-and-play addition to any post-training pipeline.

cs.LG↗

arXiv2026

FrogNano: Training a 4B Coding Agent via Online Task Synthesis

We present FrogNano, a 4B coding agent designed to tackle software engineering (SWE) tasks efficiently and effectively, even under resource-constrained environments. It is post-trained exclusively via RL on around 1,500 SWE environments with synthetic tasks. A key ingredient for improving performance is an online task synthesis pipeline that creates tasks calibrated to the frontier of learnability for the current checkpoint. This report provides evidence that competitive small coding agents can be trained with synthetic tasks alone, without traditional distillation from larger models, and that generating tasks at the learnability frontier of the current agent is important. We report details on the training methodology, evaluations across diverse environments, and in-depth analyses, serving as a foundation for our ongoing exploration of lightweight yet capable coding agents that can run on minimal hardware.

cs.AI↗

arXiv2026

SecureVibe: Making Vibe Coding More Secure

As vibe coding becomes increasingly capable and widespread, security vulnerabilities in even functionally correct solutions are a growing concern. When investigating functionally correct but insecure solutions, we find that the insecure agent is less than half as likely to conduct effective planning and testing for the hidden security risks behind the functional requirements. Motivated by this, we develop SECUREVIBE, a training recipe that explicitly targets planning and testing for code security. SECUREVIBE constructs training signals around these security behaviors. It includes supervised fine-tuning on the security suite with 4 security tasks, and post-training methods, SECUREVIBE_rl and SECUREVIBE_hg, to enhance security capabilities from verifiable execution feedback and hint-based self-supervision. Our SECUREVIBE outperforms the baseline on two types of security coding tasks across 4 benchmarks. Specifically, SECUREVIBE improves the security pass@1 by 6.9 points on BaxBench. The gains extend to unseen CWE categories, with improvements of 11.5 points on SusVibes. Meanwhile, it also improves functionality pass@1 by 13.6 points on the security coding task SusVibes and 4.1 points on the generic coding task SWE-bench Verified. Further analysis offers two practical insights: (i) diversifying supervision across security planning, coding, and testing strengthens security behaviors more effectively than adding coding trajectories alone, and (ii) hint-guided supervision is particularly valuable when the agent's existing security capabilities are insufficient to learn effectively from outcome feedback.

cs.CR↗

arXiv2026

Shockingly Simple Self-retrospection Improves Agentic Models Without RL

People learn not only by repeating successful actions, but also by recounting and explaining their experiences, revising their understanding to guide future behavior. Can a language-model agent improve its future actions by training only on explanations of its own experience? We investigate this question by studying Retrospection-Only Fine-Tuning (ROFT), a minimal online procedure designed to isolate the effect of explanation-only training on subsequent behavior. The agent attempts a task, observes available feedback, generates a retrospective explanation, and is fine-tuned with a next-token prediction loss on the explanation tokens alone. The procedure uses neither an external teacher nor a reward-based policy update. In software-engineering experiments with Qwen3.5-4B, ROFT is trained on problems with mixed successful and unsuccessful base-model attempts. On held-out SWE-bench Verified and Pro, it reaches 49.2% and 26.8% solve rates after 20 updates without using a verifier, compared with GRPO's 48.0% and 25.3% after 40 updates in the evaluated runs, and makes faster early progress in training time and sampled attempts. It also learns to solve individual tasks on which all 64 sampled base-model attempts failed, showing that learning can begin without any initially successful trajectories. Behavioral analyses find that ROFT indirectly assigns credit to actions, encouraging good actions and discouraging incorrect ones. Moreover, prompting retrospections to emphasize more direct solutions yields shorter subsequent attempts even without an explicit length penalty. Together, these findings show that learning to explain can also improve learning to do, establishing self-generated retrospections as useful training targets and motivating further study of explanation-to-action transfer.

cs.AI↗

arXiv2026

ProgramDistill: From Interactive Web Apps to Verifiable Reference-Guided SWE Tasks

Coding agents are typically evaluated with desired behavior specified through issues or instructions. In practical web development, however, agents may need to infer behavior from working software and implement it in an incomplete application. We introduce ProgramDistill, a benchmark evaluating coding agents on features discovered through interaction with fully functional reference applications. We build ProgramDistill by factorizing applications into features of different granularities, each associated with replayable behaviors executable via its gold patch. Our pipeline, mine-craft-patch, discovers 1,975 replay-verified behaviors across 26 applications and constructs 4,063 tasks without human intervention. Across nine frontier coding agents, GPT-6 Astra and Claude Opus 5 achieve 49.2% and 28.8% success on cumulative workflows in full-application reconstruction. In partial-application reconstruction, success falls from 100% to 64.0% and from 96% to 32% as restoration depth increases from 1 to 8. ProgramDistill thus provides a scalable benchmark with controlled difficulty for evaluating and diagnosing coding agents, and a natural basis for future curriculum-based training.

cs.SE↗

arXiv2026

Align to Misalign: Automatic LLM Jailbreak with Meta-Optimized LLM Judges

Identifying the vulnerabilities of large language models (LLMs) is crucial for improving their safety by addressing inherent weaknesses. Jailbreaks, in which adversaries bypass safeguards with crafted input prompts, play a central role in red-teaming by probing LLMs to elicit unintended or unsafe behaviors. Recent optimization-based jailbreak approaches iteratively refine attack prompts by leveraging LLMs. However, they often rely heavily on either binary attack success rate (ASR) signals, which are sparse, or manually crafted scoring templates, which introduce human bias and uncertainty in the scoring outcomes. To address these limitations, we introduce AMIS (Align to MISalign), a meta-optimization framework that jointly evolves jailbreak prompts and scoring templates through a bi-level structure. In the inner loop, prompts are refined using fine-grained and dense feedback from a fixed scoring template. In the outer loop, the template is optimized using an ASR alignment score, gradually evolving to better reflect true attack outcomes across queries. This co-optimization process yields progressively stronger jailbreak prompts and more calibrated scoring signals. Evaluations on AdvBench and JBB-Behaviors demonstrate that AMIS achieves state-of-the-art performance, including 88.0% ASR on Claude-3.5-Haiku and 100.0% ASR on Claude-4-Sonnet, outperforming existing baselines by substantial margins.

cs.AI↗

arXiv2026

Learning to Solve Complex Problems via Dataset Decomposition

Curriculum learning is a class of training strategies that organizes the data being exposed to a model by difficulty, gradually from simpler to more complex examples. This research explores a reverse curriculum generation approach that recursively decomposes complex datasets into simpler, more learnable components. We propose a teacher-student framework where the teacher is equipped with the ability to reason step-by-step, which is used to recursively generate easier versions of examples, enabling the student model to progressively master difficult tasks. We propose a novel scoring system to measure data difficulty based on its structural complexity and conceptual depth, allowing curriculum construction over decomposed data. Experiments on math datasets (MATH and AIME) and code generation datasets demonstrate that models trained with curricula generated by our approach exhibit superior performance compared to standard training on original datasets.

cs.LG↗

arXiv2026

MedRiskEval: Medical Risk Evaluation Benchmark of Language Models, On the Importance of User Perspectives in Healthcare Settings

As the performance of large language models (LLMs) continues to advance, their adoption in the medical domain is increasing. However, most existing risk evaluations largely focused on general safety benchmarks. In the medical applications, LLMs may be used by a wide range of users, ranging from general users and patients to clinicians, with diverse levels of expertise and the model's outputs can have a direct impact on human health which raises serious safety concerns. In this paper, we introduce MedRiskEval, a medical risk evaluation benchmark tailored to the medical domain. To fill the gap in previous benchmarks that only focused on the clinician perspective, we introduce a new patient-oriented dataset called PatientSafetyBench containing 466 samples across 5 critical risk categories. Leveraging our new benchmark alongside existing datasets, we evaluate a variety of open- and closed-source LLMs. To the best of our knowledge, this work establishes an initial foundation for safer deployment of LLMs in healthcare.

cs.CL↗

arXiv2025

Learning to Extract Context for Context-Aware LLM Inference

User prompts to large language models (LLMs) are often ambiguous or under-specified, and subtle contextual cues shaped by user intentions, prior knowledge, and risk factors strongly influence what constitutes an appropriate response. Misinterpreting intent or risks may lead to unsafe outputs, while overly cautious interpretations can cause unnecessary refusal of benign requests. In this paper, we question the conventional framework in which LLMs generate immediate responses to requests without considering broader contextual factors. User requests are situated within broader contexts such as intentions, knowledge, and prior experience, which strongly influence what constitutes an appropriate answer. We propose a framework that extracts and leverages such contextual information from the user prompt itself. Specifically, a reinforcement learning based context generator, designed in an autoencoder-like fashion, is trained to infer contextual signals grounded in the prompt and use them to guide response generation. This approach is particularly important for safety tasks, where ambiguous requests may bypass safeguards while benign but confusing requests can trigger unnecessary refusals. Experiments show that our method reduces harmful responses by an average of 5.6% on the SafetyInstruct dataset across multiple foundation models and improves the harmonic mean of attack success rate and compliance on benign prompts by 6.2% on XSTest and WildJailbreak. These results demonstrate the effectiveness of context extraction for safer and more reliable LLM inferences.

cs.LG↗

arXiv2025

BlurGuard: A Simple Approach for Robustifying Image Protection Against AI-Powered Editing

Recent advances in text-to-image models have increased the exposure of powerful image editing techniques as a tool, raising concerns about their potential for malicious use. An emerging line of research to address such threats focuses on implanting "protective" adversarial noise into images before their public release, so future attempts to edit them using text-to-image models can be impeded. However, subsequent works have shown that these adversarial noises are often easily "reversed," e.g., with techniques as simple as JPEG compression, casting doubt on the practicality of the approach. In this paper, we argue that adversarial noise for image protection should not only be imperceptible, as has been a primary focus of prior work, but also irreversible, viz., it should be difficult to detect as noise provided that the original image is hidden. We propose a surprisingly simple method to enhance the robustness of image protection methods against noise reversal techniques. Specifically, it applies an adaptive per-region Gaussian blur on the noise to adjust the overall frequency spectrum. Through extensive experiments, we show that our method consistently improves the per-sample worst-case protection performance of existing methods against a wide range of reversal techniques on diverse image editing scenarios, while also reducing quality degradation due to noise in terms of perceptual metrics. Code is available at https://github.com/jsu-kim/BlurGuard.

cs.CV↗

arXiv2025

Gistify! Codebase-Level Understanding via Runtime Execution

As coding agents are increasingly deployed in large codebases, the need to automatically design challenging, codebase-level evaluation is central. We propose Gistify, a task where a coding LLM must create a single, minimal, self-contained file that can reproduce a specific functionality of a codebase. The coding LLM is given full access to a codebase along with a specific entrypoint (e.g., a python command), and the generated file must replicate the output of the same command ran under the full codebase, while containing only the essential components necessary to execute the provided command. Success on Gistify requires both structural understanding of the codebase, accurate modeling of its execution flow as well as the ability to produce potentially large code patches. Our findings show that current state-of-the-art models struggle to reliably solve Gistify tasks, especially ones with long executions traces.

cs.CL↗

arXiv2025

BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills

High quality bugs are key to training the next generation of language model based software engineering (SWE) agents. We introduce a novel method for synthetic generation of difficult and diverse bugs. Our method instructs SWE Agents to introduce a feature into the codebase whereby they may unintentionally break tests, resulting in bugs. Prior approaches often induce an out-of-distribution effect by generating bugs intentionally (e.g. by introducing local perturbation to existing code), which does not reflect realistic development processes. We perform qualitative analysis to demonstrate that our approach for generating bugs more closely reflects the patterns found in human-authored edits. Through extensive experiments, we demonstrate that our bugs provide more efficient training data for supervised fine-tuning, outperforming other bug datasets by 2% with half the training data (1.2k vs. 3k bugs). We train on our newly generated bugs in addition to existing bug datasets to get FrogBoss a state-of-the-art 32B parameter model on SWE-bench Verified with a pass@1 of 54.6% and FrogMini a state-of-the-art 14B model on SWE-bench Verified with a pass@1 of 45.3% on SWE-bench Verified averaged over three seeds.

cs.SE↗

arXiv2025

Rethinking Safety in LLM Fine-tuning: An Optimization Perspective

Fine-tuning language models is commonly believed to inevitably harm their safety, i.e., refusing to respond to harmful user requests, even when using harmless datasets, thus requiring additional safety measures. We challenge this belief through systematic testing, showing that poor optimization choices, rather than inherent trade-offs, often cause safety problems, measured as harmful responses to adversarial prompts. By properly selecting key training hyper-parameters, e.g., learning rate, batch size, and gradient steps, we reduce unsafe model responses from 16\% to approximately 5\%, as measured by keyword matching, while maintaining utility performance. Based on this observation, we propose a simple exponential moving average (EMA) momentum technique in parameter space that preserves safety performance by creating a stable optimization path and retains the original pre-trained model's safety properties. Our experiments on the Llama families across multiple datasets (Dolly, Alpaca, ORCA) demonstrate that safety problems during fine-tuning can largely be avoided without specialized interventions, outperforming existing approaches that require additional safety data while offering practical guidelines for maintaining both model performance and safety during adaptation.

cs.LG↗

arXiv2025

Exploring Sparse Adapters for Scalable Merging of Parameter Efficient Experts

Merging parameter-efficient task experts has recently gained growing attention as a way to build modular architectures that can be rapidly adapted on the fly for specific downstream tasks, without requiring additional fine-tuning. Typically, LoRA serves as the foundational building block of such parameter-efficient modular architectures, leveraging low-rank weight structures to reduce the number of trainable parameters. In this paper, we study the properties of sparse adapters, which train only a subset of weights in the base neural network, as potential building blocks of modular architectures. First, we propose a simple method for training highly effective sparse adapters, which is conceptually simpler than existing methods in the literature and surprisingly outperforms both LoRA and full fine-tuning in our setting. Next, we investigate the merging properties of these sparse adapters by merging adapters for up to 20 natural language processing tasks, thus scaling beyond what is usually studied in the literature. Our findings demonstrate that sparse adapters yield superior in-distribution performance post-merging compared to LoRA or full model merging. Achieving strong held-out performance remains a challenge for all methods considered.

cs.LG↗

arXiv2025

Enhancing Variational Autoencoders with Smooth Robust Latent Encoding

Variational Autoencoders (VAEs) have played a key role in scaling up diffusion-based generative models, as in Stable Diffusion, yet questions regarding their robustness remain largely underexplored. Although adversarial training has been an established technique for enhancing robustness in predictive models, it has been overlooked for generative models due to concerns about potential fidelity degradation by the nature of trade-offs between performance and robustness. In this work, we challenge this presumption, introducing Smooth Robust Latent VAE (SRL-VAE), a novel adversarial training framework that boosts both generation quality and robustness. In contrast to conventional adversarial training, which focuses on robustness only, our approach smooths the latent space via adversarial perturbations, promoting more generalizable representations while regularizing with originality representation to sustain original fidelity. Applied as a post-training step on pre-trained VAEs, SRL-VAE improves image robustness and fidelity with minimal computational overhead. Experiments show that SRL-VAE improves both generation quality, in image reconstruction and text-guided image editing, and robustness, against Nightshade attacks and image editing attacks. These results establish a new paradigm, showing that adversarial training, once thought to be detrimental to generative models, can instead enhance both fidelity and robustness.

cs.LG↗

arXiv2025

debug-gym: A Text-Based Environment for Interactive Debugging

Large Language Models (LLMs) are increasingly relied upon for coding tasks, yet in most scenarios it is assumed that all relevant information can be either accessed in context or matches their training data. We posit that LLMs can benefit from the ability to interactively explore a codebase to gather the information relevant to their task. To achieve this, we present a textual environment, namely debug-gym, for developing LLM-based agents in an interactive coding setting. Our environment is lightweight and provides a preset of useful tools, such as a Python debugger (pdb), designed to facilitate an LLM-based agent's interactive debugging. Beyond coding and debugging tasks, this approach can be generalized to other tasks that would benefit from information-seeking behavior by an LLM agent.

cs.AI↗

arXiv2024

Optimizing Query Generation for Enhanced Document Retrieval in RAG

Large Language Models (LLMs) excel in various language tasks but they often generate incorrect information, a phenomenon known as "hallucinations". Retrieval-Augmented Generation (RAG) aims to mitigate this by using document retrieval for accurate responses. However, RAG still faces hallucinations due to vague queries. This study aims to improve RAG by optimizing query generation with a query-document alignment score, refining queries using LLMs for better precision and efficiency of document retrieval. Experiments have shown that our approach improves document retrieval, resulting in an average accuracy gain of 1.6%.

cs.IR↗