Search papers, labs, and topics across Lattice.
22 papers published across 2 labs.
A lineage verification method that distinguishes model ancestry with perfect accuracy, even under aggressive checkpoint modifications.
Faraday, the AI Scientist, not only outperforms leading models in replication tasks but also adopts a more rigorous scientific approach, hinting at a new era of AI-driven research.
MARC's multi-agent orchestration allows for precise clinical AI reasoning while eliminating the need for manual prompt engineering.
Non-ML cultural heritage experts can now independently analyze artifacts and validate findings using an intuitive, open-source computer vision platform.
Over 38,920 discrepancies in processor specifications were uncovered, revealing critical bugs that compromise security analysis tools.
A lineage verification method that distinguishes model ancestry with perfect accuracy, even under aggressive checkpoint modifications.
Faraday, the AI Scientist, not only outperforms leading models in replication tasks but also adopts a more rigorous scientific approach, hinting at a new era of AI-driven research.
MARC's multi-agent orchestration allows for precise clinical AI reasoning while eliminating the need for manual prompt engineering.
Non-ML cultural heritage experts can now independently analyze artifacts and validate findings using an intuitive, open-source computer vision platform.
Over 38,920 discrepancies in processor specifications were uncovered, revealing critical bugs that compromise security analysis tools.
Foundation models outperform traditional supervised methods in fall and stress detection, challenging the notion that bigger always means better in health monitoring tasks.
Access to frontier AI is becoming a critical component of national cyber defense, yet only a few states can realistically achieve true sovereignty over these technologies.
Achieving a proactive Socratic dialogue style in LLMs reveals that fine-tuning can significantly enhance cognitive flexibility and alignment across languages.
Achieving state-of-the-art results for Danish with a model that uses only permissible data challenges the notion that larger datasets are necessary for competitive performance.
Language models can autonomously resolve open mathematical conjectures at a surprisingly low cost, achieving notable success without relying on extensive prior literature.
Achieving high accuracy in domain model extraction using lightweight LLMs opens new avenues for reverse engineering in privacy-sensitive contexts.
Indian foundation models excel in traditional benchmarks but struggle with newer evaluations, revealing critical gaps in the national AI ecosystem.
Open-weight model adoption in scientific research is not just a trend toward transparency; it's a geopolitical shift, with Chinese researchers leading the charge.
A novel multivariate graph analysis reveals the hidden dynamics of predatory journals, transforming how we detect illegitimate practices in academic publishing.
The rapid rise of .cursorrules files in low-activity projects reveals a surprising gap in security considerations within AI-assisted programming prompts.
LoRA fine-tuned open-source SLMs not only surpass commercial models in critical triage tasks but also uncover high-severity cases that might otherwise go undetected.
Activation probes can uncover security vulnerabilities in AI-generated code that traditional prompting methods completely miss.
A novel reliability metric and optimization framework could save cloud services millions while enhancing user experience.
Recursive self-improvement in Macaron-V1 leads to continual learning that adapts to real-world experiences, setting a new standard for open agent models.
XFeat's accuracy-efficiency trade-off holds up under scrutiny, but its architectural claims reveal surprising nuances that could reshape future designs.
Over half of the security research artifacts fail to meet their reproducibility claims, raising serious questions about the integrity of LLM-driven vulnerability validation.
Open-weight models can achieve competitive performance in financial text comprehension, with Kimi K2.6 ranking impressively high against proprietary counterparts.
