18 papers · 1 filter
Understanding Target Probability Reuse in Speculative Decoding
Jungseob Lee, Giyoon Jang, Chailin Nam +6
Speculative decoding uses a target model to verify a tree of draft tokens in parallel. We study whether the resulting next-token distributions can help select the next tree. For un…
Target Regeneration Shrinks the Gains of Draft-Training Objectives in Speculative Decoding
Jungseob Lee, Hogun Seong, Euijun Choi +1
Speculative decoding accelerates large language models by verifying drafted tokens in parallel. New draft-training objectives are usually credited with the acceptance they add over…
MixLID: Token-Level Language Identification in Mixed-Language Text
Seongtae Hong, Youngjoon Jang, Seungyoon Lee +2
Processing multilingual text requires both identifying a broad range of languages and distinguishing the language of each token in inputs containing multiple languages. We introduc…
Predicting and Repairing Merge Collapse in Large Language Models
Jungseob Lee, Seungyoon Lee, Sugyeong Eo +3
Large language models fine-tuned from a shared base can be merged by averaging their task vectors, but some merges collapse far below the base model, and common merge operators giv…
Recovering Off-Policy Supervision for Speculative Decoding
Jungseob Lee, Chanjun Park, Sugyeong Eo +1
Block drafters for speculative decoding are commonly trained on corpora written by external models, where a single off-policy token invalidates supervision for all subsequent slots…
RenderRank: Learning to Rerank Text with Compressed Visual Tokens
Seongtae Hong, Youngjoon Jang, Jungseob Lee +2
Rendering document text as images allows vision-language models to encode documents as visual tokens, which can reduce input sequence length compared with text input. This reductio…
Where Activation Sparsity and KV-Cache Sparsity Cross in LLM Decoding
Jungseob Lee, Seungyoon Lee, Seongtae Hong +2
At each step, decoding one sequence with a large language model rereads the projection weights, whose traffic is fixed, and the key-value (KV) cache, whose traffic grows with conte…
Faster Block-Diffusion Serving with Distribution-Free Risk Guarantees
Jungseob Lee, Dongyub Jude Lee, Chanjun Park +2
Block-diffusion language models are served at hand-picked operating points, such as acceptance thresholds, buffer depth, schedule, checkpoint and precision, and each point is chose…
Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models
Jungseob Lee, Seongtae Hong, Dongyub Jude Lee +4
Speculative decoding accelerates generation without changing its output, but on vision-language models (VLMs) a self-reinforcing cycle holds it back. Because an autoregressive draf…
The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice
Jungseob Lee, Jaehyung Seo, Heuiseok Lim
Hidden-state probes effectively detect LLM hallucinations, but the geometry of the signal remains poorly characterized, driving increasingly complex probe architectures. Across thr…
Language Chain in Alignment: Cross-lingual Ranking Preference Optimization
Seungyoon Lee, Minhyuk Kim, Jungseob Lee +1
The alignment of Large Language Models heavily relies on English-centric high-quality preference data, which often leads to suboptimal performance in other languages. In this paper…
LAMAR: An Open Language-Aware Multilingual Alignment Reranker
Seongtae Hong, Youngjoon Jang, Jungseob Lee +2
In multilingual retrieval augmented generation pipelines, an embedding model can retrieve relevant documents written in multiple languages, which are subsequently reranked before a…
Answer-Conditioned Chains of Thought Degrade Verifiable-Reasoning Distillation in Large Language Models
Jungseob Lee, Seungyoon Lee, Suhyune Son +4
A standard recipe for distilling the reasoning ability of large language models (LLMs) is to sample chains of thought from the model, keep those that reach the correct final answer…
To Isolate or to Score? Model-Adaptive Assessment for Cost-Efficient Multi-Agent RAG
Jungseob Lee, Chanjun Park, Heuiseok Lim
Multi-agent document assessment for retrieval-augmented generation is computationally expensive, driving practitioners toward smaller, deployable models whose assessment mechanisms…
Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards
Jungseob Lee, Seungyoon Lee, Seongtae Hong +3
Training large language models to reason efficiently is a critical challenge. While integrating length-penalizing rewards into Group Relative Policy Optimization (GRPO) aims to red…
Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations
Dongyub Jude Lee, Jungseob Lee, Seungyoon Lee +5
Refusal on a safety benchmark does not reveal how stable that behavior will remain after model updates. Benign downstream fine-tuning can weaken refusal, yet behavioral evaluations…
DART: Draft-Agreement Routing for Training-Free Adaptive Thinking Budgets in Hybrid Reasoning Models
Jungseob Lee, Seongtae Hong, Seungjun Lee +7
Hybrid reasoning models can answer directly or spend extra tokens on extended thinking. A practical router should choose between these modes for each query, so easy problems avoid…
Improving Semantic Proximity in Information Retrieval through Cross-Lingual Alignment
Seongtae Hong, Youngjoon Jang, Jungseob Lee +2
With the increasing accessibility and utilization of multilingual documents, Cross-Lingual Information Retrieval (CLIR) has emerged as an important research area. Conventionally, C…