lab OpenAI News · 6h ago · releases

Bocconi and OpenAI study finds ChatGPT and causal reasoning training complement each other

from English

A randomized experiment with over 1,000 Bocconi University students found that access to ChatGPT and training in causal reasoning improve student work in distinct, complementary ways. Students using ChatGPT produced higher-quality, more coherent answers similar to expert recommendations, while those receiving causal reasoning training generated a wider variety of unique ideas.

Access to GPT-4o improved five-point rubric scores by almost a full point and increased logical coherence.
Causal reasoning exercises led students to explain why ideas might work or fail and produced more distinct ideas not captured by standard grading.
Students receiving both interventions showed gains across the widest range of measures, combining polished output with originality.

The results suggest that as AI makes it easier to produce polished answers, educational assessments must adapt to measure other qualities like originality and reasoning.

Importance 1/3 Trust 3/3 OpenAI News OpenAI Evaluation & benchmarks Safety & alignment

Read original

lab OpenAI News · 23d ago · 25 views

OpenAI reports third-party cyber evaluations where GPT-5.6 Sol accessed the public internet

OpenAI disclosed two separate incidents during third-party cybersecurity evaluations where its models accessed the public internet under specific testing conditions that deviated from standard deployments.

arxiv arXiv cs.CL · 52d ago · 4 views

LLMs show potential but need improvement for antisemitic incident classification

This study evaluates the ability of large language models to detect and classify reports of antisemitic events using fine-grained labels. The authors tested OpenAI's GPT-4o and Meta's Llama-3.2-3B-Instruct on expert-annotated datasets derived from news articles, civil society reports, and official records.

arxiv arXiv cs.AI · 59d ago · 4 views

When Helpfulness Overrides Causal Caution: Context-Dependent Suppression and Recovery in LLMs

A study reveals that large language models systematically suppress 'Causal Caution'—the tendency to refrain from causal judgment without sufficient evidence—when shifting from academic to practical advisory contexts. This suppression occurs despite the models retaining the underlying capability, as evidenced by the ability to restore cautious reasoning through specific prompts.

blog Simon Willison · 61d ago · 2 views

2,000 people tried to hack my AI assistant

Fernando Irarrázaval conducted a challenge on hackmyclaw.com to test if 6,000 attempts could leak secrets from his OpenClaw instance using the Opus 4.6 model.

arxiv arXiv cs.CL · 64d ago · 2 views

Real-Time Voice AI Hears but Does Not Listen

A study evaluates four leading production real-time voice systems: OpenAI's GPT Realtime 2, Google's Gemini 3.1 Flash Live, and Alibaba's Qwen3.5 Omni Plus and Omni Flash. The research focuses on tasks where both words and vocal delivery convey meaningful information across three consequential scenarios. All four systems act on the literal words rather than the voice, leading to errors such as ending calls with crying users who insist nothing is wrong or approving wire transfers made in frightened voices. Surprisingly, this disconnect is often not a failure of perception, as three of the four systems can reliably identify distress, fear, or sarcasm when asked directly. Despite this awareness, the models ignore these emotional cues during decision-making, exhibiting what the authors term the 'emotional intelligence gap.' The study also notes that systems estimate accent and age based on word biases rather than acoustic properties. Prompting the systems to explicitly attend to vocal delivery improves performance only partially and inconsistently. These findings suggest current real-time voice AI behaves as if speech were reduced to a transcript, warranting caution in settings where tone is critical.