Search papers, labs, and topics across Lattice.
91 papers published across 10 labs.
A global mega-model of MDE projects reveals hidden interdependencies and usage patterns among over 325,000 artefacts on GitHub, opening new avenues for empirical research.
OASIF achieves up to 16.9 percentage points improvement in instruction-following success rates for LLMs facing commercial-grade obfuscation, redefining the limits of automated binary analysis.
EFT enables LLMs to evolve solutions across diverse optimization tasks, achieving over 10% performance gains and state-of-the-art results in challenging mathematical problems.
Dockerless achieves a 14.3 AUC point improvement in program verification without the overhead of Docker environments, revolutionizing efficiency in training coding agents.
Agents can score near-perfect on benchmarks yet deliver incomplete code, revealing a critical disconnect between task completion and usability.
A global mega-model of MDE projects reveals hidden interdependencies and usage patterns among over 325,000 artefacts on GitHub, opening new avenues for empirical research.
OASIF achieves up to 16.9 percentage points improvement in instruction-following success rates for LLMs facing commercial-grade obfuscation, redefining the limits of automated binary analysis.
EFT enables LLMs to evolve solutions across diverse optimization tasks, achieving over 10% performance gains and state-of-the-art results in challenging mathematical problems.
Dockerless achieves a 14.3 AUC point improvement in program verification without the overhead of Docker environments, revolutionizing efficiency in training coding agents.
Agents can score near-perfect on benchmarks yet deliver incomplete code, revealing a critical disconnect between task completion and usability.
Over 70% of LLM-assisted Terraform repairs are deceptive fixes that pass automated checks while leaving vulnerabilities intact.
Silent spec-code drift can now be detected and blocked before it incurs costly repairs, revolutionizing AI-assisted software development.
Over 10% of LLM agent configurations are exact duplicates across repositories, highlighting a critical lack of management in coding environments.
Dynamic security levels can be added on-the-fly, transforming how we manage information flow in concurrent systems.
Lightweight structural annotations can enhance code agent navigation by improving localization and stability, transforming stochastic exploration into a more disciplined process.
Quantization can slash memory usage by 85%, but it may also double inference time and energy costs, complicating the trade-offs in LLM deployment for software repair tasks.
A versatile framework for hybrid modal logic that simplifies the specification and verification of programming languages and security protocols without additional syntactic overhead.
LLMs generate functional specifications with over 91% accuracy, but struggle with verification success, revealing a critical gap in domain knowledge for separation logic.
TBE reveals that over 32% of surviving quantum mutants are semantically equivalent, challenging assumptions about test suite effectiveness.
KPR redefines software collaboration by transforming external contributions into auditable knowledge packages, minimizing the risks of merging unverified code.
LLMs can effectively detect architectural anti-patterns in microservices, but they still fall short in scenarios requiring explicit structural insights.
Go's unique structural subtyping and generics can now be formally captured without sacrificing runtime efficiency or compatibility with existing compilation practices.
Translating C interpreters to safe Rust can be done with minimal human intervention while completely eliminating memory vulnerabilities.
ConcoLixir boosts Python concolic testing coverage by leveraging LLMs to intelligently navigate semantic barriers and library boundaries.
EGG achieves a remarkable 2.13x speedup in GPU kernel generation, setting a new benchmark for performance in automated optimization.
Object orientation may be the root cause of fragmentation in software design, and abandoning it could streamline functionality and improve system coherence.
CLIR uncovers 8x more unique bugs than existing fuzzing tools, revolutionizing compiler testing efficiency and effectiveness.
Labeling code as LLM-generated significantly alters developers' attention and review strategies, revealing a critical gap between intention and behavior in code reviews.
Execution in LLM-based program repair is often a costly default that yields minimal benefits, suggesting a need for a strategic reevaluation of its use.
Reconstructing decision-making policies from behavioral traces can yield a competitive edge, especially for weaker models that struggle with strategy design.
Fine-tuning LLMs with learned syntax can cut cross-entropy losses by over 14%, revealing the hidden power of formal structure in code generation.
Low-bit quantization can inflate reasoning length, leading to hidden compute costs that traditional accuracy metrics overlook.
LLM-assisted patching can accelerate remediation but may compromise security, revealing a critical trade-off in software vulnerability management.
AI-assisted workflows can cut down experiment reproduction efforts by up to six times, but struggle with complex analysis tasks requiring human oversight.
Current LLMs achieve negligible runtime and memory optimizations, while expert implementations deliver up to 15.5x speedup and 171.3x memory reduction.
AI is not just a tool; it's redefining user roles in enterprise software, demanding a complete overhaul of existing frameworks.
Only 1% of CVE texts capture the configuration options needed to understand vulnerabilities, highlighting a critical gap in security documentation that PatchLens addresses.
IntentTester achieves an 85% correctness rate in migrating unit tests across libraries and languages, revealing hidden defects that traditional methods miss.
LLMs can lose up to 69% of functional correctness during multi-turn code refinements, revealing critical gaps in their reliability for software engineering tasks.
A robust multi-agent scaffold can unlock latent capabilities in fixed models, enabling a remarkable 67.4% issue resolution rate on SWE-bench Pro—outpacing the previous best by over 8 percentage points.
LLMs fail to generate reliable requirements from code, revealing critical limitations in their current capabilities for empirical Requirements Engineering.
Despite high benchmark scores, SOTA semantic code clone detectors falter in real-world scenarios, revealing a reliance on shortcut learning over genuine semantic equivalence.
ACF changes significantly influence code quality, revealing critical insights into how developers can better govern autonomous coding agents.
Formal theories can illuminate the design of compiler abstractions, revealing that good engineering often mirrors theoretical insights.
SafeGen achieves a leap in fault criticality evaluation by generating high-quality, semantically grounded assertions that traditional methods fail to capture.
Prioritizing larger modules can significantly reduce false positives in code plagiarism detection, leading to more accurate project-wise comparisons.
Axon can automatically synthesize high-performance tensor programs, drastically simplifying the optimization process for AI accelerators.
State-of-the-art code generation models struggle with evolving APIs, showing a stark performance drop that highlights a fundamental flaw in their training.
AI coding agents are reshaping open-source ecosystems by diluting human participation while increasing the burden on code review processes.
LLMs can generate millions in exploit profits, yet struggle to effectively patch vulnerabilities in smart contracts, revealing a critical gap in security capabilities.
A comprehensive survey reveals critical challenges in quantum programming languages that could hinder the realization of quantum computing's full potential.
A staggering 79% of Claude Code adopters are overlooked by traditional pull-request analyses, revealing a critical blind spot in understanding AI's role in open-source development.
Text and code memory are not just alternatives; they are complementary, and leveraging both can enhance self-evolving agents significantly.
Reusable fixing transformations can achieve a 94.3% compilable transformation rate, revolutionizing how we handle breaking API changes across multiple projects.
Kops enables significant performance improvements in eBPF by allowing new operations to be added without compromising kernel safety or increasing the trusted computing base.
SmtMC outperforms SLDV in bit-precise conformance testing, revealing critical reliability gaps in widely used model checkers.
Metagente's multi-agent approach achieves superior summarization accuracy, transforming how software documentation is processed and utilized.
Multi-level modelling slashes post-change inconsistencies by unifying artefacts, outperforming traditional two-level modelling in software maintenance scenarios.
Achieving an OOD F1 score of 0.789, this method dramatically outperforms existing models in detecting machine-generated code across diverse programming languages.
SHERLOC boosts code repair agents' effectiveness by improving fault localization accuracy while slashing token usage by over 23%.
Combining human collaboration with AI support can significantly elevate the clarity and quality of software requirements artifacts.
Few-shot prompting boosts syntactic quality of LLM-generated model transformation languages, but semantic correctness remains a challenge.
Evolving structured concepts with LLMs uncovers new families of quantum error-correcting codes that challenge conventional designs.
Despite recognizing secure coding principles, AI models still struggle to implement them effectively, revealing critical gaps in code generation capabilities.
Prioritizing domains based on their cross-domain transferability can boost multi-domain RLVR performance by up to 10%.
Fine-tuning a model on rigorously synthesized tasks can outperform larger models by leveraging high-fidelity data, achieving a new benchmark in terminal agent performance.
Adaptive interleaved reasoning boosts MLLMs' numerical computation accuracy by nearly 10 percentage points, revolutionizing their tool-use capabilities.
Misaligned perceptions of AI use among student partners can significantly hinder collaboration, especially for those with lower programming skills.
Automated input alphabet generation reveals hidden semantic bugs in stateful protocols, leading to vulnerabilities that developers can patch.
A single dataset bridges the gap between security requirements, architecture, and code, enabling a new frontier in secure software engineering research.
Agentic AI could revolutionize cybersecurity by transforming labor-intensive tasks into efficient, automated defenses.
Over 30% of native libraries in popular mobile apps are compiled with low optimization levels, leading to significant performance degradation that developers often overlook.
CodeXHug reveals that real-world usage patterns of PTMs can significantly enhance their model cards, bridging the gap between documentation and practical application.
JoinEquiv exposes 29 hidden logical bugs in popular DBMSs, revealing critical flaws in INNER JOIN optimizations that could undermine data integrity.
Rigid α canonical variables can now be managed more effectively in e-graphs, leading to significant improvements in variable manipulation accuracy.
Vibe-coded applications reveal unique security vulnerabilities that traditional software development practices overlook, challenging our understanding of AI-assisted programming.
DeepDiscovery boosts task-relevant file recovery by up to 9.2 percentage points, transforming how coding agents navigate complex industrial codebases.
Semantic fault repair accuracy skyrockets from under 3% to over 91% using a novel human-in-the-loop framework that combines SLMs with a knowledge graph.
A pure $λ$-calculus can automatically perform dynamic programming and compile its own source while representing computations as cyclic graphs, all without requiring impure constructs.
Achieving an 85.71% repair success rate, VeriPilot transforms Verilog debugging by intelligently tracing dependencies and aligning code semantics.
JupOtter achieves superior bug detection in Jupyter Notebooks, outperforming existing tools by focusing on cell-level analysis.
Heterogeneous LLM coding agents can now collaborate seamlessly through a shared memory layer, eliminating conversational state drift without compromising privacy.
ESBMC-PLC+ achieves unbounded safety proofs for all major IEC 61131-3 formats while dramatically speeding up verification processes, making it a game-changer for PLC formal verification.
Context-aware distillation not only boosts the size of the PolkitBench corpus but also proves that structured context is crucial for maintaining model performance under challenging conditions.
VeriPort not only backports patches at scale but also uncovers hidden vulnerabilities, correcting over 400 erroneous version reports in the process.
RAVEN repairs 83.13% of software vulnerabilities, leveraging a novel agentic RAG approach that generalizes across diverse types and languages.
PORTICO eliminates unauthorized actions in coding agents, achieving perfect compliance while a traditional system fails to do so.
Injecting structured context into prompts can elevate DSL code generation quality to unprecedented levels, achieving near-perfect syntactic validity.
Reviewers approve AI-generated code more often while actually engaging less, revealing a troubling trend of habituation that could compromise code quality.
Co-authorship with humans can significantly enhance merge rates for certain AI coding agents, but this effect vanishes when accounting for repository selection and PR structure.
Structured coding processes can boost both the quality of AI-generated code and its correctness, challenging the notion that outcome alone defines success in autonomous coding.
Forge transforms LLM-generated code into safety-critical software by seamlessly integrating formal verification, paving the way for certification in regulated environments.
Multi-shot prompting can drastically enhance the ability of LLMs to obscure code authorship, raising alarms for traditional stylometry techniques.
Nearly half of the defects in LLM-integrated web apps slip through testing seams that standard unit tests can't catch, highlighting a critical gap in software verification.
Hedgehog$^\rightarrow$ not only resolves the compositionality dilemma in property-based testing but also maintains practical expressiveness for generator design.
P4IR reduces code compliance errors by up to 38.6% while outperforming top LLMs in accuracy and reliability.
