RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
super hub Canonical reference
Code Llama: Open Foundation Models for Code
Canonical reference. 80% of citing Pith papers cite this work as background.
316
Pith papers citing it
398
external citations · Pith
Background
80% of classified citations
abstract
We release Code Llama, a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models, infilling capabilities, support for large input contexts, and zero-shot instruction following ability for programming tasks. We provide multiple flavors to cover a wide range of applications: foundation models (Code Llama), Python specializations (Code Llama - Python), and instruction-following models (Code Llama - Instruct) with 7B, 13B, 34B and 70B parameters each. All models are trained on sequences of 16k tokens and show improvements on inputs with up to 100k tokens. 7B, 13B and 70B Code Llama and Code Llama - Instruct variants support infilling based on surrounding content. Code Llama reaches state-of-the-art performance among open models on several code benchmarks, with scores of up to 67% and 65% on HumanEval and MBPP, respectively. Notably, Code Llama - Python 7B outperforms Llama 2 70B on HumanEval and MBPP, and all our models outperform every other publicly available model on MultiPL-E. We release Code Llama under a permissive license that allows for both research and commercial use.
hub tools
citation-role summary
background 39
method 4
baseline 3
dataset 2
other 1
citation-polarity summary
claims ledger
- abstract We release Code Llama, a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models, infilling capabilities, support for large input contexts, and zero-shot instruction following ability for programming tasks. We provide multiple flavors to cover a wide range of applications: foundation models (Code Llama), Python specializations (Code Llama - Python), and instruction-following models (Code Llama - Instruct) with 7B, 13B, 34B and 70B parameters each. All models are trained on sequences of 16k tokens and show improvements on inputs with up
authors
co-cited works
representative citing papers
The first SoK on LLM-based AutoPT frameworks provides a six-dimension taxonomy of agent designs and a unified empirical benchmark evaluating 15 frameworks via over 10 billion tokens and 1,500 manually reviewed logs.
The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.
CodeTracer attributes backdoored code completions to responsible fine-tuning examples via structured behavioral fingerprints, embedding retrieval, and LLM semantic reasoning under realistic post-deployment constraints.
UI2App introduces a benchmark showing that vision-language models can reconstruct web page visuals but largely fail to infer the underlying interaction logic from screenshots alone.
A training-free post-hoc code watermark embeds 24-bit identifiers via multi-channel spread-spectrum encoding over naming conventions and semantic pattern pairs, with majority voting and Reed-Solomon recovery.
BOUND refines LLMs' package-validity boundary via targeted editing to cut package hallucination rates by 79.9% on edit prompts and 65.4% on unseen prompts in recommendation tasks while generalizing to code generation.
FirmCure uses LLM-based adaptive perception, reflective synthesis, and autonomous runtime intervention to rehost 21 firmware images at 100% port opening and 90.5% interactivity rates across 5 architectures.
MetaDataGenAgent is a multimodal multi-agent system with a closed-loop plan-execute-reflect mechanism that autonomously extracts, validates, and encodes metamaterial data from unstructured literature into executable databases.
Incomplete constrainers in constrained decoding push LLMs into low-probability program regions, making unconstrained decoding outperform constrained decoding on functional correctness across seven models and three benchmarks.
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.
SENTINEL generates targeted tasks from model failures in a Controller-Proposer-Solver loop, raising Pass^1 from 66.4 to 74.9 on Tau2-Bench Retail and outperforming standard RL.
Authors demonstrate functional memorization in code LLMs via counterfactual midtraining comparison on functional equivalence metrics beyond textual overlap.
Strong coding agents use metaprogramming to solve tasks in unfamiliar esoteric languages while weaker agents do not, with performance gaps larger than in mainstream benchmarks.
OpenRTLSet supplies 131k+ Verilog samples with AI-generated descriptions to enable fine-tuning of LLMs for hardware module design.
Introduces the binning semiring and causal graphical models to show that correlational evaluation of learnability in formal language tasks leads to incorrect conclusions from confounders.
PrivCode++ introduces the first DP code generation method protecting both prompts and code via latent-conditioned two-stage training, claiming higher utility and stronger privacy than prior baselines.
Multilingual execution-grounded benchmark finds top open code LLM at 23.64% correctness versus 57.2% human baseline, with compile errors dominating 63% of failures.
SkelDPO improves code generation efficiency by 2-7% over prior DPO methods via joint preference losses on full code and efficiency-critical skeletons.
Structurally distinct circuits for literal sequence copying across token frequency bands implement the same computation, shown by broad transfer of band-specific edges, a shared core recovering 99% performance, and interchangeable representations via causal interventions.
ADK Arena evaluates 51 Python ADKs by having an LLM learn each framework's API, write and repair agent code, and run on benchmarks, finding 57% success rate, 5.6x cost variation, no dominant framework, and substitutable information sources.
3DCodeBench is a new benchmark evaluating 12 VLMs on translating multimodal prompts into procedural 3D modeling code, paired with 3DCodeArena for human preference rankings.
Hybrid vector-search plus fingerprinting pipeline for LLM code provenance achieves Winnowing-level MRR on short snippets and up to 5.4% better on longer ones at logarithmic query time.
CopT reverses CoT by eliciting a draft answer first then using continuous-embedding contrastive verification and on-policy thinking to reflect and correct, yielding up to 23% higher accuracy and 57% fewer tokens without training.
citing papers explorer
-
Efficient Training on Multiple Consumer GPUs with RoundPipe
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
-
Hackers or Hallucinators? A Comprehensive Analysis of LLM-Based Automated Penetration Testing
The first SoK on LLM-based AutoPT frameworks provides a six-dimension taxonomy of agent designs and a unified empirical benchmark evaluating 15 frameworks via over 10 billion tokens and 1,500 manually reviewed logs.
-
Why Do Multi-Agent LLM Systems Fail?
The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.
-
Beware What You Autocomplete: Forensic Attribution of Backdoored Code Completions
CodeTracer attributes backdoored code completions to responsible fine-tuning examples via structured behavioral fingerprints, embedding retrieval, and LLM semantic reasoning under realistic post-deployment constraints.
-
UI2App: Benchmarking Visual Interaction Inference in Executable Web Application Generation
UI2App introduces a benchmark showing that vision-language models can reconstruct web page visuals but largely fail to infer the underlying interaction logic from screenshots alone.
-
Multi-Channel Spread-Spectrum Code Watermarking
A training-free post-hoc code watermark embeds 24-bit identifiers via multi-channel spread-spectrum encoding over naming conventions and semantic pattern pairs, with majority voting and Reed-Solomon recovery.
-
Mitigating Package Hallucinations in Large Language Models via Model Editing
BOUND refines LLMs' package-validity boundary via targeted editing to cut package hallucination rates by 79.9% on edit prompts and 65.4% on unseen prompts in recommendation tasks while generalizing to code generation.
-
FirmCure:Towards Autonomous and Adaptive Rehosting of Linux-Based Firmware
FirmCure uses LLM-based adaptive perception, reflective synthesis, and autonomous runtime intervention to rehost 21 firmware images at 100% port opening and 90.5% interactivity rates across 5 architectures.
-
Autonomous Generation of Metamaterial Databases Based on Multimodal Agents
MetaDataGenAgent is a multimodal multi-agent system with a closed-loop plan-execute-reflect mechanism that autonomously extracts, validates, and encodes metamaterial data from unstructured literature into executable databases.
-
The Alignment Problem in Constrained Code Generation
Incomplete constrainers in constrained decoding push LLMs into low-probability program regions, making unconstrained decoding outperform constrained decoding on functional correctness across seven models and three benchmarks.
-
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.
-
SENTINEL: Failure-Driven Reinforcement Learning for Training Tool-Using Language Model Agents
SENTINEL generates targeted tasks from model failures in a Controller-Proposer-Solver loop, raising Pass^1 from 66.4 to 74.9 on Tau2-Bench Retail and outperforming standard RL.
-
Detecting Functional Memorization in Code Language Models
Authors demonstrate functional memorization in code LLMs via counterfactual midtraining comparison on functional equivalence metrics beyond textual overlap.
-
Frontier Coding Agents Use Metaprogramming to Adapt to Unfamiliar Programming Languages
Strong coding agents use metaprogramming to solve tasks in unfamiliar esoteric languages while weaker agents do not, with performance gaps larger than in mainstream benchmarks.
-
OpenRTLSet: A Fully Open-Source Dataset for Large Language Model-based Verilog Module Design
OpenRTLSet supplies 131k+ Verilog samples with AI-generated descriptions to enable fine-tuning of LLMs for hardware module design.
-
Causally Evaluating the Learnability of Formal Language Tasks
Introduces the binning semiring and causal graphical models to show that correlational evaluation of learnability in formal language tasks leads to incorrect conclusions from confounders.
-
PrivCode++: Latent-Conditioned Differentially Private Code Generation for Comprehensive Guarantees
PrivCode++ introduces the first DP code generation method protecting both prompts and code via latent-conditioned two-stage training, claiming higher utility and stronger privacy than prior baselines.
-
Beyond Pass Rate: A Multilingual, Execution-Grounded Evaluation of Open Code LLMs
Multilingual execution-grounded benchmark finds top open code LLM at 23.64% correctness versus 57.2% human baseline, with compile errors dominating 63% of failures.
-
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.
-
Many Circuits, One Mechanism: Input Variation and Evaluation Granularity in Circuit Discovery
Structurally distinct circuits for literal sequence copying across token frequency bands implement the same computation, shown by broad transfer of band-specific edges, a shared core recovering 99% performance, and interchangeable representations via causal interventions.
-
ADK Arena: Evaluating Agent Development Kits via LLM-as-a-Developer
ADK Arena evaluates 51 Python ADKs by having an LLM learn each framework's API, write and repair agent code, and run on benchmarks, finding 57% success rate, 5.6x cost variation, no dominant framework, and substitutable information sources.
-
3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code
3DCodeBench is a new benchmark evaluating 12 VLMs on translating multimodal prompts into procedural 3D modeling code, paired with 3DCodeArena for human preference rankings.
-
Efficient and Scalable Provenance Tracking for LLM-Generated Code Snippets
Hybrid vector-search plus fingerprinting pipeline for LLM code provenance achieves Winnowing-level MRR on short snippets and up to 5.4% better on longer ones at logarithmic query time.
-
CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning
CopT reverses CoT by eliciting a draft answer first then using continuous-embedding contrastive verification and on-policy thinking to reflect and correct, yielding up to 23% higher accuracy and 57% fewer tokens without training.
-
BacktestBench: Benchmarking Large Language Models for Automated Quantitative Strategy Backtesting
Introduces BacktestBench benchmark with 18k QA pairs across four backtesting tasks and evaluates 23 LLMs via the AutoBacktest multi-agent system.
-
Constrained Code Generation with Discrete Diffusion
Constrained Diffusion for Code (CDC) integrates constraint satisfaction into the reverse denoising process of discrete diffusion models via constraint-aware operators that use optimization and program analysis to steer generation toward feasible programs.
-
Hydra: Efficient, Correct Code Generation via Checkpoint-and-Rollback Support
Hydra enables asynchronous static error checking and targeted checkpoint-rollback repair during LLM code generation, cutting latency by up to 71% and token use by up to 70% versus post-hoc repair on C/C++ tasks.
-
Reward-Weighted On-Policy Distillation with an Open Property-Equivalence Verifier for NL-to-SVA Generation
Reward-Weighted On-Policy Distillation with an open property-equivalence verifier produces a 7B model that surpasses prior SOTA on NL-to-SVA generation across pass@1/5/10 metrics.
-
LGMT: Logic-Grounded Metamorphic Testing for Evaluating the Reasoning Reliability of LLMs
LGMT is a logic-grounded metamorphic testing framework that detects hidden reasoning defects in LLMs by checking consistency on semantically invariant inputs derived from FOL equivalences.
-
SmartEval: A Benchmark for Evaluating LLM-Generated Smart Contracts from Natural Language Specifications
SmartEval is a new benchmark showing LLM-generated smart contracts score 8.29 points higher than expert versions on average but frequently omit logic (35.3%) or mishandle state transitions (23.4%).
-
MeshFIM: Local Low-Poly Mesh Editing via Fill-in-the-Middle Autoregressive Generation
MeshFIM enables local low-poly mesh editing by autoregressively filling target regions conditioned on context, using boundary markers, positional embeddings, and a gated geometry encoder to enforce attachment, topology, and region limits.
-
Mean-Pooled Cosine Similarity is Not Length-Invariant: Theory and Cross-Domain Evidence for a Length-Invariant Alternative
Mean-pooled cosine similarity grows with sequence length in anisotropic transformer embeddings independent of content, while CKA shows far less length dependence across code, translation, and vision tasks.
-
Evaluating Non-English Developer Support in Machine Learning for Software Engineering
Code LLMs generate substantially worse comments outside English, and no tested automatic metric or LLM judge reliably matches human assessment of those outputs.
-
Delta-Based Neural Architecture Search: LLM Fine-Tuning via Code Diffs
Fine-tuned 7B LLMs generating unified diffs for neural architecture refinement achieve 66-75% valid rates and 64-66% mean first-epoch accuracy, outperforming full-generation baselines by large margins while cutting output length by 75-85%.
-
Coral: Cost-Efficient Multi-LLM Serving over Heterogeneous Cloud GPUs
Coral cuts multi-LLM serving costs by up to 2.79x and raises goodput by up to 2.39x on heterogeneous GPUs through adaptive joint optimization and a lossless two-stage decomposition that solves quickly.
-
QASecClaw: A Multi-Agent LLM Approach for False Positive Reduction in Static Application Security Testing
A multi-agent LLM system cuts false positives in static application security testing by 88.6% on the OWASP Benchmark while dropping recall by only 3.1%.
-
VulKey: Automated Vulnerability Repair Guided by Domain-Specific Repair Patterns
VulKey introduces hierarchical expert knowledge abstractions to guide LLMs in vulnerability repair, reporting 31.5% accuracy on PrimeVul (7.6% above best baseline) and strong results on Vul4J.
-
Social Bias in LLM-Generated Code: Benchmark and Mitigation
LLMs show up to 60.58% social bias in generated code; a new Fairness Monitor Agent cuts bias by 65.1% and raises functional correctness from 75.80% to 83.97%.
-
When Prompt Under-Specification Improves Code Correctness: An Exploratory Study of Prompt Wording and Structure Effects on LLM-Based Code Generation
Structurally rich task descriptions make LLMs robust to prompt under-specification, and under-specification can enhance code correctness by disrupting misleading lexical or structural cues.
-
Constraint-Guided Multi-Agent Decompilation for Executable Binary Recovery
A constraint-guided multi-agent system turns raw decompiler output into re-executable code at 84-97% success rates, outperforming prior LLM decompilation methods on real binaries.
-
PhysCodeBench: Benchmarking Physics-Aware Symbolic Simulation of 3D Scenes via Self-Corrective Multi-Agent Refinement
PhysCodeBench benchmark and SMRF multi-agent framework enable better AI generation of physically accurate 3D simulation code, boosting performance by 31 points over baselines.
-
RAG-Reflect: Agentic Retrieval-Augmented Generation with Reflections for Comment-Driven Code Maintenance on Stack Overflow
RAG-Reflect achieves F1=0.78 on valid comment-edit prediction using retrieval-augmented reasoning and self-reflection, outperforming baselines and approaching fine-tuned models without retraining.
-
Assessing the Impact of Requirement Ambiguity on LLM-based Function-Level Code Generation
Orchid benchmark shows requirement ambiguity degrades LLM code generation performance across all models, with advanced models hit hardest, and LLMs rarely detect or resolve the ambiguity themselves.
-
Parallel-SFT: Improving Zero-Shot Cross-Programming-Language Transfer for Code RL
Parallel-SFT mixes parallel programs across languages during SFT to produce more transferable RL initializations, yielding better zero-shot generalization to unseen programming languages.
-
IRIS: Interpolative R\'enyi Iterative Self-play for Large Language Model Fine-Tuning
IRIS unifies self-play fine-tuning under an interpolative Rényi objective with adaptive alpha scheduling and reports better benchmark scores than baselines while surpassing full supervised fine-tuning with only 13% of the annotated data.
-
PlayCoder: Making LLM-Generated GUI Code Playable
PlayCoder raises the rate of LLM-generated GUI apps that can be played end-to-end without logic errors from near zero to 20.3% Play@3 by adding repository-aware generation, agent-driven testing, and iterative repair.
-
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.
-
Efficient Low-Resource Language Adaptation via Multi-Source Dynamic Logit Fusion
TriMix dynamically fuses logits from three model sources to outperform baselines and Proxy Tuning on eight low-resource languages across four model families.
-
Structural Anchors and Reasoning Fragility:Understanding CoT Robustness in LLM4Code
CoT prompting in LLM4Code shows mixed robustness that depends on model family, task structure, and perturbations destabilizing structural anchors, leading to trajectory deformations like lengthening, branching, and simplification.
-
CodeComp: Structural KV Cache Compression for Agentic Coding
CodeComp uses Joern-extracted Code Property Graph priors for training-free structural KV cache compression, outperforming attention-only baselines on bug localization and code generation while matching full-context patch quality.