REVIEW 19 cited by
CodeGemma: Open Code Models Based on Gemma
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
This paper introduces CodeGemma, a collection of specialized open code models built on top of Gemma, capable of a variety of code and natural language generation tasks. We release three model variants. CodeGemma 7B pretrained (PT) and instruction-tuned (IT) variants have remarkably resilient natural language understanding, excel in mathematical reasoning, and match code capabilities of other open models. CodeGemma 2B is a state-of-the-art code completion model designed for fast code infilling and open-ended generation in latency-sensitive settings.
Forward citations
Cited by 19 Pith papers
-
Incentivizing Vision Language Models to Search for Long Video Question Answering
RL post-training of a VLM agent with neuro-symbolic temporal-logic rewards for evidence retrieval raises Pass@1 by up to 8% and Pass@4 by 15% on long-video QA.
-
CrossPL: Evaluating Large Language Models on Cross Programming Language Code Generation
CrossPL, a 1,982-task benchmark built from GitHub repositories, shows that LLMs achieve at most 79.74% pass@1 on cross-language IPC code generation and struggle with low-level protocols like Pipe.
-
Zero-Shot Detection of LLM-Generated Code via Approximated Task Conditioning
ATC detects AI-generated code by asking a language model to reconstruct the programming task, then scoring token entropy under that reconstructed task, outperforming prior zero-shot detectors on Python, C++, and Java ...
-
Structure-Aware Fill-in-the-Middle Pretraining for Code
AST-FIM masks complete syntax-tree subtrees during fill-in-the-middle pretraining, improving infilling performance on real-world code edits.
-
Doc2CI: A Multi-Service Study of CI Configuration Generation Using Large Language Models
Across four CI services and 16 LLMs, exact reproduction of reference configs stays below 3.2%, while schema validity (about 71%) trails parse rate (97%), showing similarity and usability are different objectives.
-
Towards Reliable C-to-Rust Translation with Rule-Guided Reasoning and Reinforcement Learning
A rule-guided MCTS plus dual-reward reinforcement learning pipeline improves LLM-based C-to-Rust translation accuracy and cuts unsafe Rust output across three benchmarks.
-
Functional and Secure Code Generation with Task Vectors
LPO-derived Secure-Anchored task vectors raise simultaneous functional-and-secure code rates by 2.1–36 pp on six coding LLMs with near-zero inference overhead.
-
Prompt Orchestration Markup Language
POML is a markup language that structures LLM prompts, embeds multimodal data, and decouples formatting via stylesheets, with case studies showing strong prompt format sensitivity.
-
AutoBridge: Automating Smart Device Integration with Centralized Platform
AutoBridge generates platform-compliant IoT integration code with 93.87% success and 94.87% function coverage in fully automatic mode, reaching 100% coverage with minimal yes/no human feedback.
-
Can LLMs Write CI? A Study on Automatic Generation of GitHub Actions Configurations
Across six LLMs, zero-shot generation of GitHub Actions YAML configurations reaches 69% similarity to ground truth but only 3% exact matches, with code-pretrained models performing no better than general-purpose ones.
-
Zero-Shot Attribution for Large Language Models: A Distribution Testing Approach
Anubis re-frames LLM attribution as a distribution testing problem with EVAL access, and reports AUROC above 0.9 on code benchmarks with around 2000 samples, beating detectGPT.
-
RedCoder: Automated Multi-Turn Red Teaming for Code LLMs
A multi-turn red-teaming agent trained on simulated attacker-defender conversations induces vulnerable code at higher rates than prior attack methods across several code LLMs.
-
SV-TrustEval-C: Evaluating Structure and Semantic Reasoning in Large Language Models for Source Code Vulnerability Analysis
Large language models perform poorly on a new C-code vulnerability benchmark, indicating they rely on pattern matching rather than genuine reasoning.
-
A Qualitative Investigation into LLM-Generated Multilingual Code Comments and Automatic Evaluation Metrics
Neural metrics for evaluating code comments are unreliable for multilingual output, often scoring random noise as high as real generated comments.
-
CLOVER: A Test Case Generation Benchmark with Coverage, Long-Context, and Verification
CLOVER is a new Python unit-test-generation benchmark with 845 problems, contexts up to 128k tokens, and coverage-based retrieval, where all 14 evaluated models score below 35% on the hardest task.
-
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.
-
Efficient Code LLM Training via Distribution-Consistent and Diversity-Aware Data Selection
Selecting 10K code instructions with a parametric feature-space model lifts DeepSeekCoder-Base-6.7B from 67.1% to 69.5% on HumanEval and from 74.9% to 77.2% on MBPP versus full 92K training, in single-run evaluations.
-
GeoPandas-AI: A Smart Class Bringing LLM as Stateful AI Code Assistant
A new open-source Python class, GeoDataFrameAI, adds a stateful LLM chat interface directly to GeoPandas data frames for geospatial code generation and analysis.
-
Reinforcing Code Generation: Improving Text-to-SQL with Execution-Based Learning
Execution-based reinforcement learning raises SQLCoder-7B's exact-match accuracy from 31.49% to 49.83% on TEMPTABQA-C, approaching SQLCoder-70B, but shows no gain for CodeGemma.
Discussion (0). Continue with ORCID to comment.