REVIEW 56 cited by
Magicoder: Empowering Code Generation with OSS-Instruct
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
We introduce Magicoder, a series of fully open-source (code, weights, and data) Large Language Models (LLMs) for code that significantly closes the gap with top code models while having no more than 7B parameters. Magicoder models are trained on 75K synthetic instruction data using OSS-Instruct, a novel approach to enlightening LLMs with open-source code snippets to generate diverse instruction data for code. Our main motivation is to mitigate the inherent bias of the synthetic data generated by LLMs through the wealth of open-source references for the production of more realistic and controllable data. The orthogonality of OSS-Instruct and other data generation methods like Evol-Instruct further enables us to build an enhanced MagicoderS. Both Magicoder and MagicoderS substantially outperform state-of-the-art code models with similar or even larger sizes on a wide range of coding benchmarks. Notably, MagicoderS-CL-7B based on CodeLlama even surpasses the prominent ChatGPT on HumanEval+ (66.5 vs. 65.9 in pass@1 ). Overall, OSS-Instruct opens a new direction for crafting diverse synthetic instruction data for code using abundant open-source references.
Forward citations
Cited by 56 Pith papers
-
Large Language Diffusion Models
LLaDA is a scalable diffusion-based language model that matches autoregressive LLMs like LLaMA3 8B on tasks and surpasses GPT-4o on reversal poem completion.
-
Repository-Level Solidity Code Generation with Large Language Models: From Prompting to Fine-Tuning
Introduces SolidityBench benchmark and SolidityScore metric for repository-level Solidity code generation, finding supervised fine-tuning outperforms prompting, CoT, ICL, and RAG methods on evaluated LLMs.
-
CODEBLOCK: Learning to Supervise Code at the Right Granularity
CodeBlock partitions code responses into syntactically coherent blocks, scores them with generalized cross-entropy and data-flow signals, and applies sparse supervision to achieve higher pass@1 than full SFT using 1.9...
-
SkelDPO: A Skeleton-Guided Direct Preference Optimization Framework for Efficient Code Generation
SkelDPO improves code generation efficiency by 2-7% over prior DPO methods via joint preference losses on full code and efficiency-critical skeletons.
-
RogueMerge: Robust and Unified Attacks against LLM Model Merging
RogueMerge is a unified attack method that jointly optimizes task vectors to succeed after merging, using stochastic min-max simulation for unknown merging settings and a Taylor-approximated DRO for prompt generalizat...
-
Reconstruction of Personally Identifiable Information from Supervised Finetuned Models
PII can be reconstructed from SFT models via prefix attacks, with the new COVA algorithm improving success rates and leakage varying by attacker knowledge and PII type.
-
Beyond Translation Accuracy: Addressing False Failures in LLM-Based Code Translation
Many reported failures in LLM-based code translation are false negatives due to evaluation pipeline issues such as improper compilation flags, missing library links, and unconfigured runtime environments rather than i...
-
Cascaded Code Editing: Large-Small Model Collaboration for Effective and Efficient Code Editing
A cascaded large-small model system generates edit sketches with the large model and applies them with the small model to make code editing both accurate and token-efficient.
-
SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios
SWE-EVO shows GPT-5.4 with OpenHands reaching only 25% success on complex multi-file evolution tasks versus 72.8% on SWE-Bench Verified, and introduces Fix Rate as a partial-progress metric.
-
Assessing Coherency and Consistency of Code Execution Reasoning by Large Language Models
LLMs achieve 81% coherent execution simulation on HumanEval but show mostly random or weak consistency across tests, with frontier models relying on natural language shortcuts instead of true program analysis.
-
rStar2-Agent: Agentic Reasoning Technical Report
A 14B model trained with agentic RL and a resample-on-correct rollout strategy scores 80.6% on AIME24 and 69.8% on AIME25, nearly matching DeepSeek-R1 (671B) in one week on 64 GPUs.
-
Towards Agentic Runtime Healing
Healer uses LLMs to dynamically generate and execute runtime error-handling code, with GPT-4 recovering from 72.8% of errors across four datasets.
-
LEAP: Lean Environment-Feedback via Adaptive Pruning for Code RL in GPU Kernel Generation
LEAP combines difficulty-conditioned pruning with pairwise rank-based rewards to improve multi-turn reinforcement learning for CUDA kernel generation.
-
MoE$^2$-LoRA: When MoE Models Meet MoE-style Low-Rank Adaptation
Routing LoRA adapters with the frozen base router's logits plus a shared cross-layer adapter pool gives the best PEFT accuracy and retention on three MoE backbones.
-
SciCodePile: A 128GB Corpus and Executable Benchmark for Challenging Scientific Code Generation
SciCodePile contributes a 128GB scientific code corpus and a 200-task executable benchmark showing that current LLMs still fall far short of reliable scientific code generation.
-
Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models
Mid-training on function-aware fill-in-the-middle data improves SWE-Bench by 2.8-5.4 points and reduces post-training capability loss.
-
Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models
Function-aware FIM mid-training on ordinary code improves coding-agent SWE-Bench scores and reduces post-training capability erosion on non-agent benchmarks.
-
Are we Merging the Right Models? Impact of Expert Training Duration on Model Merging for LLMs
The optimal training duration for expert models before merging depends on the merging method; sparsification-based merging benefits from overtrained experts.
-
LP-SFT: Local-Preserving Supervised Fine-Tuning via Multimodal Entropy Structure
A fine-tuning objective that preserves the base model's local preference among non-label tokens improves pass@1 accuracy and retains more pretrained capability than standard SFT.
-
Test Case Selection for Deep Neural Networks: A Replication Study on LLMs for Code
Replication of TCS strategies on 17 LLM instances across three code tasks shows only partial generalization from vision DNN results, with uncertainty features aiding early failure discovery and representation features...
-
Lost in the Flow with Code Talkers: Unveiling the Instruction-Tuning Tax of Large Language Models in Code Tasks
Empirical study finds instruction tuning on CodeLLMs improves instruction following at the expense of infilling performance, termed the Instruction-Tuning Tax.
-
BenchEvolver: Frontier Task Synthesis via Solution-Centric Evolution
BenchEvolver evolves coding problem solutions to generate harder, valid tasks, producing LiveCodeBench-Plus where frontier models score 27.5-62.6% and enabling RL gains on held-out tests.
-
TRACER: A Semantic-Aware Framework for Fine-Grained Contamination Detection in Code LLMs
TRACER presents a semantic-aware framework and the first benchmark for fine-grained code contamination detection across three levels of overlap, reporting F1 scores of 0.91-0.92 and large gains over prior methods.
-
Asking Back: Interaction-Layer Antidistillation Watermarks
Interaction-layer antidistillation watermarks use system-prompt-induced behavioral markers like explicit follow-up questions that transfer to distilled student models at 45-89% relative fidelity and can be audited via...
-
Bayesian Model Merging
Bayesian Model Merging introduces a bi-level optimization framework that merges task-specific models via closed-form Bayesian regression with an anchor prior and global hyperparameter search, outperforming baselines a...
-
GLAN-QnA-KR: A Seedless Taxonomy-Driven Korean Instruction Corpus
A 303,581-row Korean instruction corpus generated seedlessly from a 1,084-discipline taxonomy, with near-zero duplicates and low measured overlap with KMMLU, KoBEST, and HAE-RAE-Bench.
-
Optimizer-Model Consistency: Full Finetuning with the Same Optimizer as Pretraining Forgets Less
Full finetuning with the pretraining optimizer reduces forgetting compared to other optimizers or LoRA while achieving comparable new-task performance.
-
Adversarial Arena: Crowdsourcing Data Generation through Interactive Competition
Adversarial competition between attacker and defender teams generates diverse multi-turn conversational data that improves LLM performance on secure code generation benchmarks by 18-29%.
-
Sensitivity-Positional Co-Localization in GQA Transformers
In Llama 3.1 8B, task-sensitive layers cluster late while RoPE adaptation is strongest early, yet applying both adaptations only to sensitivity-identified layers outperforms other layer choices by 4-16 points on MMLU,...
-
Self-Supervised Bootstrapping of Action-Predictive Embodied Reasoning
R&B-EnCoRe uses self-supervised importance-weighted variational inference to distill action-predictive reasoning datasets that improve VLA performance on manipulation, navigation, and driving tasks without external verifiers.
-
Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts
EMoE trains MoE models so they maintain performance when the number of activated experts changes at inference, expanding the usable range to 2-3 times the training k with higher peak results.
-
Modeling Code: Is Text All You Need?
A GNN-encoded LLVM IR graph, prepended as soft prompts to a frozen code LLM, improves accuracy on device mapping, algorithm classification, vulnerability detection, and code translation tasks.
-
Turning the Tide: Repository-based Code Reflection
A new 1,888-case, six-language benchmark and a trained model measure and improve how code LLMs edit and repair code inside multi-file repositories.
-
BlockFFN: Towards End-Side Acceleration-Friendly Mixture-of-Experts with Chunk-Level Activation Sparsity
A ReLU-routed MoE with chunk-level sparsity training objectives and custom kernels combining activation sparsity with speculative decoding achieves over 70% 8-token chunk sparsity and up to 3.67x end-side speedup.
-
Multilingual Multimodal Software Developer for Code Generation
A 7B vision-language model trained on synthetic diagram-to-code data outperforms several larger open-weight models on a new 10-language UML/flowchart code-generation benchmark.
-
Does Learning Mathematical Problem-Solving Generalize to Broader Reasoning?
Long chain-of-thought and RL training on math problems improves general reasoning benchmarks, while short chain-of-thought math fine-tuning often degrades performance.
-
A Hierarchical and Evolvable Benchmark for Fine-Grained Code Instruction Following with Multi-Turn Feedback
MultiCodeIF introduces a 2,021-task, 14-language benchmark with 27 constraint types to evaluate code instruction following, finding that multi-level constraints sharply reduce model success and iterative feedback subs...
-
Agentless: Demystifying LLM-based Software Engineering Agents
Agentless, a basic three-phase LLM pipeline for bug localization, repair, and validation, outperforms complex open-source agents on SWE-bench Lite with 32% success rate at $0.70 cost.
-
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
MiniCPM 1.2B and 2.4B models reach parity with 7B-13B LLMs via model wind-tunnel scaling and a WSD scheduler that yields a higher optimal data-to-model ratio than Chinchilla scaling.
-
LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
LiveCodeBench collects 400 recent contest problems to create a contamination-free benchmark evaluating LLMs on code generation and related capabilities like self-repair and execution.
-
Are we Merging the Right Models? Impact of Expert Training Duration on Model Merging for LLMs
Sparsification-based model merging peaks well past expert validation optima, while averaging degrades with overtraining; duration and merge method must be chosen jointly.
-
Route-Align-Verify for Functional Correctness in Code Generation
RAV combines task-aware prompt routing, aligned LoRA fine-tuning, and public-test-based candidate selection to improve MBPP pass@1, but its verification step leaks evaluation tests.
-
Quantize with Confidence? An Empirical Study of Quantization for Code Generation
4-bit quantization of code LLMs mostly preserves pass@1, but AQLM matches or beats full precision while QuIP# degrades most on complex prompts; sensitivity is model-dependent.
-
LP-SFT: Local-Preserving Supervised Fine-Tuning via Multimodal Entropy Structure
Preserving adaptive non-label local structure from the base model during SFT improves the pass@1 vs pass@k trade-off and reduces catastrophic forgetting versus vanilla cross-entropy and recent SFT variants.
-
ROSUM-MCTS: Monte Carlo Tree Search-Inspired HDL Code Summarization with Structural Rewards
ROSUM-MCTS applies MCTS-inspired hierarchical candidate expansion and a composite reward balancing functional correctness, local content adequacy, and fluency to improve LLM summaries of VHDL and Verilog code, outperf...
-
Lossless Anti-Distillation Sampling
LADS is a sampling method that keeps benign user generations statistically identical to the original model while forcing correlated samples across a distiller's multiple accounts, provably worsening their generalizati...
-
Beyond Translation Accuracy: Addressing False Failures in LLM-Based Code Translation
A large-scale study finds that many LLM code translation failures are false negatives due to improper evaluation configurations rather than incorrect translations.
-
A Survey on Evaluating Quality and Trustworthiness in LLM-Generated Data
A metric-oriented survey that classifies intrinsic quality and trustworthiness metrics for LLM-generated data across six modalities and documents systematic evaluation gaps in the current literature.
-
SelfAug: Mitigating Catastrophic Forgetting in Retrieval-Augmented Generation via Distribution Self-Alignment
Adding a KL penalty between fine-tuned and original model logits on input tokens during RAG fine-tuning reduces catastrophic forgetting while preserving downstream performance.
-
Grove MoE: Towards Efficient and Superior MoE LLMs with Adjugate Experts
Grove MoE uses unequal-size adjugate experts with complexity-based activation to run 33B-parameter models at roughly 3.1 to 3.3B active parameters while matching larger open models in benchmarks.
-
Training-Free versus Training-Based Intent Classification in LLMs: Accuracy, Robustness, and Failure Modes
Statistical classifiers built on LLM activation norms and coordinates match or beat trained MLP heads on coarse intent routing and resist camouflage better, while MLPs win on fine-grained subfield distinctions.
-
Reward-Free Code Alignment from Pretrained or Fine-Tuned LLM: Unpacking the Trade-offs for Code Generation
Empirical study on five LLMs finds pretrained-to-aligned paths yield bigger gains over baseline than finetuned-to-aligned paths, though absolute accuracy remains lower for pretrained starts.
-
Context-Instrumental Data Distillation for Kubernetes Manifest Generation: Method and Experimental Evaluation
Context-instrumental data distillation allows a 1.5B SLM to generate valid Kubernetes manifests at 91.5% pass@1 rate, with strict output formatting proving more impactful than additional training data.
-
Large Language Models for Multilingual Code Intelligence: A Survey
A survey of methods, benchmarks, and open challenges for large language models in multilingual code generation and translation.
-
Position: Intelligent Coding Systems Should Write Programs with Justifications
A position paper advocating that intelligent coding systems should accompany code with justified explanations that are cognitively aligned and semantically faithful.
-
A Survey on Large Language Models for Code Generation
A systematic literature review that organizes recent work on LLMs for code generation into a taxonomy covering data curation, model advances, evaluations, ethics, environmental impact, and applications, with benchmark...
Discussion (0). Sign in to comment.