qualia-lab
Behavioral evaluation framework for sentience-, emotion-, and welfare-related AI claims, with anti-sandbagging analysis.
- 1
- Python
- MIT
- 8 mo ago
codernate92
Exploring cross-domain AI risk, AI governance, and LLM-enhanced threat intelligence
Built from codernate92's public GitHub profile, Research Triangle, NC. Not on Take My Repo yet, so there are no posts or updates here.
Behavioral evaluation framework for sentience-, emotion-, and welfare-related AI claims, with anti-sandbagging analysis.
Threat intelligence platform with IOC ingestion, LLM summarization, and analyst workflow support.
Working paper codebase and artifact bundle for a systems-dynamics and POMG-DSD framework for frontier AI risk.
Adversarial red-teaming framework for LLM agents — ATT&CK-inspired taxonomy, 39 attack probes, adaptive campaign execution, CVSS-analog vulnerability scoring, and professional red-team report generation
Systemic-leverage AI red teaming: Meadows leverage points + STAMP/STPA + Weidinger sociotechnical layers
Mechanistic model diffing of a narrow JSON-format fine-tune in a tiny causal transformer
Python-first reproduction of the public MONA Camera Dropbox experiments
Empirical AI scaling analysis — scaling law fitting, capability emergence detection, compute trend forecasting, and publication-quality visualizations from curated frontier model data
Cyber threat monitoring dashboard with geospatial visualization, interactive charts, and analyst-oriented reporting.
Threat-hunting automation framework integrating OSQuery, CloudTrail, TheHive, and Splunk.
Horizon-Eval: evaluation-integrity framework for portable long-horizon agent benchmarks, with QA gates, trajectory auditing, replayable run bundles, and safety-gap analysis.
Multi-agent orchestration and reliability research stack with DAG scheduling, async execution, critic evaluation, and trajectory analysis.
Model evaluation and behavioral auditing framework for LLM agents — detects specification gaming, deceptive alignment, sandbagging, and corrigibility failures
Auditable training-data safety filtering stack for toxicity, PII, bias, poisoning, and robustness evaluation.
Experimental Python router for dispatching requests across multiple LLM providers and task types.
Go CLI for mapping threat actors to malware, infrastructure, and MITRE ATT&CK techniques.
Critical-infrastructure threat modeling with spatial SQL, geospatial visualization, and open-source intelligence data.
Experimental multimodal research assistant for local document ingestion, embedding, and question answering workflows.
Exploratory discourse-analysis notebooks and text analysis experiments.