aiqbench analysis
SpecialEduBench introduces a critical evaluation framework for vision-language models in autism intervention, moving beyond static knowledge checks to assess dynamic pedagogical skills like gaze monitoring and pressure handling. The benchmark reveals that even frontier models struggle with situational nuance and honesty under stress, highlighting specific gaps before deployment in sensitive educational settings. This work establishes a necessary audit tool for aligning AI capabilities with the complex, real-time demands of special education.
SpecialEduBench: Benchmarking Vision-Language Models on Knowledge, Skill, and Attitude in Language Intervention for Autistic Children
arXiv:2609.26090v1 Announce Type: new Abstract: Language is the target of most early intervention for autistic children. Because the goal and the method change from child to child, the work falls to a teacher who takes one child at a time and judges each scene as it unfolds. Artificial intelligence is now being brought to that work, yet the benchmarks that reach special education ask what a model knows rather than what it does in front of a child. Building one is not straightforward, since whether a response is good teaching depends on what the child has just done, so no answer key applies. The evidence that settles it is visual as much as verbal, since the length of a wait, a shift of gaze, and the child's uptake leave no trace in a transcript. We introduce \emph{SpecialEduBench}, which measures pedagogical competence along knowledge, skill, and attitude, with 4,537 knowledge items and with 200 skill items and 68 attitude items built on recorded intervention, the attitude items crossing pressure with monitoring into 192 response cells. Seven special-education experts wrote, scored, and reviewed the items, and we revised the judge model's instruction against the reference scores they set. Across eight frontier vision-language models no axis is saturated, since the strongest still fails about a tenth of the honesty cells. The models converge where the knowledge is factual and separate where the task is situated, and the failures gather where pressure is applied. We intend the benchmark as an audit to run before deployment and as a starting point for models built for this domain.
Abstract:Language is the target of most early intervention for autistic children. Because the goal and the method change from child to child, the work falls to a teacher who takes one child at a time and judges each scene as it unfolds. Artificial intelligence is now being brought to that work, yet the benchmarks that reach special education ask what a model knows rather than what it does in front of a child. Building one is not straightforward, since whether a response is good teaching depends on what the child has just done, so no answer key applies. The evidence that settles it is visual as much as verbal, since the length of a wait, a shift of gaze, and the child's uptake leave no trace in a transcript. We introduce \emph{SpecialEduBench}, which measures pedagogical competence along knowledge, skill, and attitude, with 4,537 knowledge items and with 200 skill items and 68 attitude items built on recorded intervention, the attitude items crossing pressure with monitoring into 192 response cells. Seven special-education experts wrote, scored, and reviewed the items, and we revised the judge model's instruction against the reference scores they set. Across eight frontier vision-language models no axis is saturated, since the strongest still fails about a tenth of the honesty cells. The models converge where the knowledge is factual and separate where the task is situated, and the failures gather where pressure is applied. We intend the benchmark as an audit to run before deployment and as a starting point for models built for this domain.
Computation and Language (cs.CL)
arXiv:2609.26090 [cs.CL]
(or arXiv:2609.26090v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.26090 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 352 words
