Source-linked AI summary
Think in English, Answer in Korean: Efficient Adaptation of Multilingual Tool-Using Agents
Utsav Garg, Sungjin Hong, Jason Jung, Justin Lee, Shaan Desai, Joon Hee Kim, Anirudh Shrinivason, Edmond Wen, Susie Park
TL;DR
Korean-English enterprise agents must combine private-data reasoning, tool use, language-consistent answers, and constrained memory and serving budgets. The model is adapted from fully post-trained Command A using preamble-conditioned hybrid SFT, RLVR, language-consistency rewards, preference alignment, and 4-bit quantization. Verifiable NL2SQL rewards, language-consistency penalties, and preference alignment improve tool use while preserving Korean and English instruction-following behavior; 4-bit quantization enables single-H100 deployment with little observed quality loss.
Problem
Korean-English enterprise agents must combine private-data reasoning, tool use, language-consistent answers, and constrained memory and serving budgets.
Method
The model is adapted from fully post-trained Command A using preamble-conditioned hybrid SFT, RLVR, language-consistency rewards, preference alignment, and 4-bit quantization.
Results
Verifiable NL2SQL rewards, language-consistency penalties, and preference alignment improve tool use while preserving Korean and English instruction-following behavior; 4-bit quantization enables single-H100 deployment with little observed quality loss.
Takeaways & Limitations
Adapting an existing post-trained model toward verifiable enterprise workflows can combine focused agent capabilities with preserved general behavior and lower serving requirements.
Takeaways & Limitations
Some evaluations are internal and proprietary data and evaluation details are limited, constraining reproducibility and requiring public benchmarks to complement deployment-specific tests.
Abstract
from arXiv · showhide
We present LuckyStar 111B, a 111B-parameter hybrid reasoning model developed through a collaboration between Cohere and LG CNS for Korean-English enterprise agents under practical memory and serving constraints. The model trains from Cohere's fully post-trained Command A model rather than a new pretraining run, and uses preamble conditioning to switch between concise non-reasoning behavior and longer tool-oriented reasoning. We study four choices for scaling tool-using agents efficiently: multilingual supervised fine-tuning, reinforcement learning with verifiable rewards for multi-step tool-use tasks, language-consistency rewards for Korean user-facing responses, and 4-bit quantization for single-GPU serving. The adapted model improves mathematical reasoning, function calling, and agentic natural-language-to-SQL (NL2SQL) performance while preserving general Korean and English instruction-following quality. These results provide a practical recipe and failure-mode analysis for adapting post-trained multilingual models to verifiable agentic workflows under memory-constrained deployment.
1. Introduction
LuckyStar 111B adapts a fully post-trained multilingual model for Korean-English enterprise agents that must combine reasoning, tool use, and constrained deployment. Its hybrid design adds specialized capabilities while preserving general instruction following and supporting language-consistent, memory-efficient operation.
- LuckyStar 111B adapts Cohere’s fully post-trained Command A for Korean-English enterprise agents instead of training a new model from scratch.The approach targets agent quality, verifiable tool use, and memory footprint together.
- Preamble conditioning lets one set of weights switch between longer reasoning for tools, mathematics, and NL2SQL and concise ordinary responses.Downstream systems select reasoning depth at inference time without separate models.
- The three-stage adaptation pipeline combines hybrid SFT, RLVR for verifiable reasoning and tool use, and preference alignment for concise user-facing behavior.This adds reasoning and tool-use behavior without full retraining.
- A language-consistency reward penalty addresses Korean prompts drifting toward English final answers during RLVR.The intervention targets multilingual reasoning failures while retaining English-heavy intermediate reasoning.
- 4-bit quantization preserves benchmark quality in the reported evaluations while enabling single-H100 serving for the 111B-parameter model.This reduces deployment requirements for enterprise settings.
2. Method
The method adapts a post-trained multilingual model through hybrid SFT, verifiable-reward reinforcement learning, and preference alignment for reasoning and tool use. It addresses multilingual reasoning imbalance, constructs executable NL2SQL supervision, and corrects RLVR-induced language drift and verbosity.
- Hybrid SFT: Hybrid SFT teaches reasoning and non-reasoning modes, using roughly 80% reasoning examples and 20% instruction-following examples.Preambles select longer intermediate reasoning or concise final responses without changing weights or architecture.
- Multilingual Reasoning Strategy: English reasoning outperformed Korean reasoning on Korean prompts in early experiments, motivating a mixed-language strategy with English traces and Korean final answers.Machine-translating 30% of traces into Korean produced little improvement, so 20,000 Korean prompts were used for verified mixed-language examples.
- Agentic NL2SQL Data: Executable NL2SQL examples are created by retaining successful gold queries with outputs between one and 500 tokens, yielding 100,000 verifiable prompt-query pairs.Executable outputs support automated checking for RLVR and rejection sampling.
- Agentic NL2SQL Data: Because the base policy achieved less than 1% accuracy on the curated NL2SQL set, best-of-N rejection-sampled SFT bootstrapped the policy before RLVR.The RLVR mixture contains over 50,000 agentic NL2SQL prompts and 4,000 mathematics prompts, with binary verifier or judge rewards.
- RLVR and Language Consistency: A language-consistency penalty reduces Korean-to-English answer drift by checking whether final responses match prompt language while reasoning remains in English.Prompt languages are pre-classified and ambiguous prompts are excluded.
- RLVR and Language Consistency: RLVR improves reasoning and agentic NL2SQL while reducing repetition failures, but it also increases verbosity and harms brief-response instruction following.Rollouts over 32k tokens are filtered rather than assigned zero reward to avoid teaching premature termination.
- Offline Preference Alignment: Low-learning-rate DPO uses alignment and Korean-specific preference pairs to restore conciseness without compromising reasoning and tool-use capabilities acquired during RLVR.Direct answers are preferred over verbose post-RLVR outputs.
3. Results
LuckyStar 111B는 수학 추론, 함수 호출, 기업용 NL2SQL에서 Command A보다 향상되었으며, 일반적인 한국어·영어 성능과 4비트 배포 효율을 함께 유지했다.
- 3.1. Reasoning and Tool Use: LuckyStar 111B는 Command A보다 한국어·영어 수학 추론, BFCL v3 함수 호출, 내부 기업용 도구 사용 평가에서 향상되었다.수학 평가는 AIME 2024와 MATH 500의 영어 및 한국어 번역 버전을 포함한다.
- 3.3. Efficient Deployment: 내부 평가는 배포별 워크플로와 스키마를 반영하지만, 도구 사용 결과는 4비트 모델에 대해 보고되지 않았다.4비트 도구 사용 평가가 최종 평가 시점까지 완료되지 않았기 때문이다.
- 3.1. Reasoning and Tool Use: Enterprise NL2SQL 점수는 7.3에서 38.0으로, LG Agentic Evaluation 점수는 2.67에서 4.85로 향상되었다.절대 점수는 현실적인 기업용 도구 사용이 여전히 어렵다는 점도 보여준다.
- 3.2. General Instruction Following: 한국어 성능은 KMMLU 68.6, ARC Challenge 89.2, IFEval 77.8, MT-Bench 8.09에 도달했고, 영어 성능은 각각 82.7, 93.8, 89.4, 8.45였다.한국어 ARC, IFEval, MT-Bench는 영어 평가의 번역 버전이다.
- 3.3. Efficient Deployment: 4비트 양자화는 메모리 사용량을 약 50% 줄이고 단일 80GB H100에서 111B 모델을 실행하게 하면서 보고된 벤치마크 품질을 FP8과 가깝게 유지했다.이는 완전한 서빙 연구가 아니라 배포 가능성 점검이며, 지연시간·처리량·에너지 측정은 남은 과제다.
4. Discussion
논의는 영어 내부 추론과 한국어 사용자 응답을 분리하는 설계, NL2SQL의 콜드스타트 필요성, 그리고 배포 중심 적응의 실용성을 강조한다.
- 4. Discussion: 영어 추론 흔적은 검증 가능한 추론에 더 강한 학습 신호를 제공하고, 언어 일관성 보상은 한국어 프롬프트의 영어 최종 답변을 줄인다.그러나 한국어 고유 추론은 문화적·도메인 특화 지식이 중요한 작업에서 여전히 미래 과제다.
- 4. Discussion: NL2SQL에서는 거부 샘플링 SFT가 RL이 정책을 개선할 만큼의 초기 능력을 제공하므로, 콜드스타트 데이터가 필요하다.이는 보상 품질과 프롬프트 선택이 검증 가능한 에이전트 RL의 핵심이라는 관찰과 일치한다.
- 4. Discussion: 개발 중 확인된 NL2SQL 콜드스타트, 한국어-영어 언어 이동, 추론 최적화 후 장황함은 각각 SFT, 언어 보상, 선호 정렬로 다뤄졌다.저자들은 이를 검증 가능한 보상, 다국어 행동, 배포 지향 적응의 접점에서 발생하는 문제로 본다.
- 4. Discussion: 배포 중심 관점에서는 모든 능력에 최적화된 단일 모델보다, 기존 사후학습 모델을 검증 가능한 워크플로에 맞추고 양자화하는 방식이 유용할 수 있다.이 방식은 일반 지시 따르기를 보존하면서 서빙 요구량을 줄이는 것을 목표로 한다.
- 4. Discussion: 일부 평가는 내부 워크플로와 제한된 독점 데이터에 의존하므로, 공개 벤치마크를 대체하기보다 보완하며 재현성도 제한된다.법적·개인정보·기밀성 의무가 독점 데이터와 내부 평가 세부사항 공개를 제한한다.
5. Conclusion
LuckyStar 111B combines hybrid reasoning, tool-use improvements, and language-consistency training while preserving bilingual instruction following. Four-bit quantization enables deployment on a single H100 with little observed quality loss.
- 4-bit quantization makes LuckyStar 111B deployable on a single H100 with little observed quality loss in the reported evaluations.
- Preamble conditioning lets one model support both reasoning and non-reasoning modes.
- Verifiable NL2SQL rewards, language-consistency penalties, and preference alignment improve tool use while preserving Korean and English instruction-following behavior.
- Future work targets native-language reasoning data, Korean domain adaptation, comprehensive serving-efficiency measurement, and closer integration with enterprise knowledge bases and tools.
LLM/Agent Usage Disclosure
The paper used LLMs throughout model development and also for proofreading, while retaining author responsibility for technical claims, experiments, and final text.
- LLMs generated candidate reasoning traces, judged answer equivalence when deterministic verification was insufficient, checked final-answer language consistency, and produced candidate SQL traces with a read-only execution tool.
- LLMs also helped proofread the paper for clarity and flow, while the authors remained responsible for technical claims, experiments, and the final text.