Six state-of-the-art LLMs systematically prefer Standard American English over AAE continuations, and a training-free activation steering method reduces this bias 5-20x more than prompting while preserving fluency.
hub Canonical reference
International AI Safety Report
Canonical reference. 83% of citing Pith papers cite this work as background.
16
Pith papers citing it
Background
83% of classified citations
abstract
The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.
hub tools
citation-role summary
background 6
citation-polarity summary
roles
background 6representative citing papers
RedVox benchmark shows speech model safety and fairness vulnerabilities persist under non-adversarial conditions, worsen in non-English languages, and increase with spoken inputs.
TukaBench extends JailbreakBench to African languages via human translation, cultural adaptation, curated prompts, and code-switching, finding lower refusal rates for culturally grounded prompts and surfacing comprehension and judging limitations.
Introduces Unlearning Depth Score (UDS) via activation patching to quantify LLM unlearning depth and claims it outperforms 20 other metrics in faithfulness and robustness on 150 models.
Proposes affective safety as a distinct class of AI harms with a taxonomy of self-alienation, bias, and relational harms, arguing that existing safety frameworks address it narrowly or not at all and calling for dedicated approaches focused on cumulative and identity-level effects.
Annotator Policy Models learn safety policies from labeling behavior alone, accurately predicting responses and revealing sources of disagreement like policy ambiguity and value pluralism.
Analysis of Canada's Federal AI Register reveals it frames AI as reliable internal tooling by obscuring sociotechnical elements like human discretion, turning transparency into performative compliance.
The 2025 AI Agent Index catalogs technical and safety details for 30 deployed AI agents and finds low developer transparency on safety, evaluations, and societal impacts.
Twelve frontier AI safety frameworks score between 8% and 34% on adapted risk-management criteria, with a median of 18%, leaving them too vague to serve as reliable external accountability mechanisms.
An accurate and trusted AI system cannot achieve human-level reasoning because there exist tasks easily solvable by humans but not by the system.
Deep feedforward ReLU networks generalize two-layer principles via paths, piecewise linear manifolds, and continuity restriction to explain training solutions.
Empirical study across multiple benchmarks finds the link between uncertainty estimators and LLM hallucinations is highly variable and often weak.
LLMs generally fail to maintain stable worldviews under adversarial conversational pressure, indicating they lack core beliefs akin to those in human cognition.
The paper unifies perspectives on Long CoT in reasoning LLMs by introducing a taxonomy, detailing characteristics of deep reasoning and reflection, and discussing emergence phenomena and future directions.
The paper analyzes evolving security and safety threats in generative AI from content generation to agentic actions, noting that attack surfaces expand faster than defenses and that many safeguards require institutional coordination not yet in place.
AGI may arrive by 2030-2040 and reshape global power balances, requiring Europe to close gaps in compute, talent retention, industrial adoption, and unified policy responses through a coordinated preparedness agenda.
citing papers explorer
-
LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering
Six state-of-the-art LLMs systematically prefer Standard American English over AAE continuations, and a training-free activation steering method reduces this bias 5-20x more than prompting while preserving fluency.
-
RedVox: Safety and Fairness Gaps in Speech Models Across Languages
RedVox benchmark shows speech model safety and fairness vulnerabilities persist under non-adversarial conditions, worsen in non-English languages, and increase with spoken inputs.
-
TukaBench: A Culturally Grounded Jailbreak Benchmark for African Languages
TukaBench extends JailbreakBench to African languages via human translation, cultural adaptation, curated prompts, and code-switching, finding lower refusal rates for culturally grounded prompts and surfacing comprehension and judging limitations.
-
Measuring the Depth of LLM Unlearning via Activation Patching
Introduces Unlearning Depth Score (UDS) via activation patching to quantify LLM unlearning depth and claims it outperforms 20 other metrics in faithfulness and robustness on 150 models.
-
Affective AI Safety: The Missing Piece in LLM Safety
Proposes affective safety as a distinct class of AI harms with a taxonomy of self-alienation, bias, and relational harms, arguing that existing safety frameworks address it narrowly or not at all and calling for dedicated approaches focused on cumulative and identity-level effects.
-
Understanding Annotator Safety Policy with Interpretability
Annotator Policy Models learn safety policies from labeling behavior alone, accurately predicting responses and revealing sources of disagreement like policy ambiguity and value pluralism.
-
Bureaucratic Silences: What the Canadian AI Register Reveals, Omits, and Obscures
Analysis of Canada's Federal AI Register reveals it frames AI as reliable internal tooling by obscuring sociotechnical elements like human discretion, turning transparency into performative compliance.
-
The 2025 AI Agent Index: Documenting Technical and Safety Features of Deployed Agentic AI Systems
The 2025 AI Agent Index catalogs technical and safety details for 30 deployed AI agents and finds low developer transparency on safety, evaluations, and societal impacts.
-
Evaluating AI Providers' Frontier Safety Frameworks
Twelve frontier AI safety frameworks score between 8% and 34% on adapted risk-management criteria, with a median of 18%, leaving them too vague to serve as reliable external accountability mechanisms.
-
Limitations on Accurate, Trusted, Human-level Reasoning
An accurate and trusted AI system cannot achieve human-level reasoning because there exist tasks easily solvable by humans but not by the system.
-
On the Principles of Deep Feedforward ReLU Networks
Deep feedforward ReLU networks generalize two-layer principles via paths, piecewise linear manifolds, and continuity restriction to explain training solutions.
-
Evaluating the Relevance of Uncertainty Estimators for LLM Hallucination
Empirical study across multiple benchmarks finds the link between uncertainty estimators and LLM hallucinations is highly variable and often weak.
-
Do LLMs have core beliefs?
LLMs generally fail to maintain stable worldviews under adversarial conversational pressure, indicating they lack core beliefs akin to those in human cognition.
-
Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models
The paper unifies perspectives on Long CoT in reasoning LLMs by introducing a taxonomy, detailing characteristics of deep reasoning and reflection, and discussing emergence phenomena and future directions.
-
From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI
The paper analyzes evolving security and safety threats in generative AI from content generation to agentic actions, noting that attack surfaces expand faster than defenses and that many safeguards require institutional coordination not yet in place.
-
Europe and the Geopolitics of AGI: The Need for a Preparedness Plan
AGI may arrive by 2030-2040 and reshape global power balances, requiring Europe to close gaps in compute, talent retention, industrial adoption, and unified policy responses through a coordinated preparedness agenda.