paper-with-me

Papers

Embarrassingly Simple Self-Distillation Improves Code Generation

2026-04-01 · Ruixiang Zhang, Richard He Bai, Huangjie Zheng, Navdeep Jaitly, Ronan Collobert, Yizhe Zhang arxiv

Can a large language model (LLM) improve at code generation using only its own raw outputs, without a verifier, a teacher model, or reinforcement learning? We answer in the affirmative with simple self-distillation (SSD): sample solutions from the model with certain temperature and truncation configurations, then fine-tune on those samples with standard supervised fine-tuning. SSD improves Qwen3-30B-Instruct from 42.4% to 55.3% pass@1 on LiveCodeBench v6, with gains concentrating on harder problems, and it generalizes across Qwen and Llama models at 4B, 8B, and 30B scale, including both instruct and thinking variants. To understand why such a simple method can work, we trace these gains to a precision-exploration conflict in LLM decoding and show that SSD reshapes token distributions in a context-dependent way, suppressing distractor tails where precision matters while preserving useful diversity where exploration matters. Taken together, SSD offers a complementary post-training direction for improving LLM code generation. Our code is available at https://github.com/apple/ml-ssd

📄 PDF Abstract BibTeX arXiv:2604.01193

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement LearningCode Generation

Results from the Paper

RankTaskDatasetModelMetrics
#134 Code Generation LiveCodeBench Embarrassingly Simple Self-Distillation Pass@1: 55.3

Similar Papers 제목 키워드 기반

Few-Medoids: An Embarrassingly Simple Coreset Selection Method for Few-Shot Knowledge Distillation

2026-07-07 · Cemil-Andrei Dilmac, Florinel-Alin Croitoru, Radu Tudor Ionescu arxiv

Coreset selection aims to identify a small and highly representative subset of a massive dataset for efficient model training. The problem remains challenging even in the few-shot knowledge distillation (KD) setup, where…

Knowledge DistillationImage Classification

Just CHOP: Embarrassingly Simple LLM Compression

2023-05-24 · Ananya Harsh Jha, Tom Sherborne, Evan Pete Walsh, Dirk Groeneveld 외

Large language models (LLMs) enable unparalleled few- and zero-shot reasoning capabilities but at a high computational footprint. A growing assortment of methods for compression promises to reduce the computational burde…

Knowledge DistillationLanguage ModelingLanguage ModellingLarge Language Model+1

An Embarrassingly Simple Backdoor Attack on Self-supervised Learning

2022-10-13 · ICCV 2023 1 · Changjiang Li, Ren Pang, Zhaohan Xi, Tianyu Du 외

As a new paradigm in machine learning, self-supervised learning (SSL) is capable of learning high-quality representations of complex data without relying on labels. In addition to eliminating the need for labeled data, r…

Adversarial RobustnessBackdoor AttackSelf-Supervised Learning

Sliced-Wasserstein Autoencoder: An Embarrassingly Simple Generative Model

2018-04-05 · Soheil Kolouri, Phillip E. Pope, Charles E. Martin, Gustavo K. Rohde

In this paper we study generative modeling via autoencoders while using the elegant geometric properties of the optimal transport (OT) problem and the Wasserstein distances. We introduce Sliced-Wasserstein Autoencoders (…

model

Sliced-Wasserstein Autoencoder: An Embarrassingly Simple Generative Model

2018-04-05 · Anonymous

In this paper we study generative modeling via autoencoders while using the elegant geometric properties of the optimal transport (OT) problem and the Wasserstein distances. We introduce Sliced-Wasserstein Autoencoders (…

model