HYDERABAD: An ordinary photograph of a room, combined with a written instruction, could help artificial intelligence create an animated human movement. This is one of five research projects from the International Institute of Information Technology, Hyderabad (IIIT‑H), selected for NeurIPS 2026, a leading international AI conference.
The project, ENACT, led by Prof. Charu Sharma, uses existing AI tools to understand a photograph, plan an action and generate the corresponding movement. “Since casual images of indoor environments exist aplenty, we leveraged those along with free‑form text instruction to create animated motion,” Sharma said. The technology could help generate training material for robots, although it is currently limited to static scenes and cannot handle actions such as opening drawers or carrying objects.
Another paper, developed with Adobe Research, addresses the difficulty of controlling individual parts of AI‑generated images. It allows users to specify details such as an animal’s left front leg rather than treating the entire animal as one subject. “Part‑level control is still an open problem for image generation models like Nano Banana2, Flux2.0 & co,” said Harsh Rangwani of Adobe Research.
The other three studies look at making AI systems fairer and more efficient. Safe‑Fair MACPO, by Ankita Kushwaha, Kiran Ravish and Preeti Lamba under Prof. Pawann Kumar, tracks whether AI agents carry more than their fair share of work through a measure called “fairness debt”, while prioritising safety. FedUCA, led by Prof. Sai Kiran Bulusu with researchers from IISc, addresses organisations dropping out of collaborative AI training by adjusting their contributions according to the benefits they expect.
TernaryLM, by Prof. Priyesh Shukla and Nisharg Nargund, is a 132‑million‑parameter language model that uses only three parameter values: ‑1, 0 and +1. It used 2.4 times less memory than an equivalent 32‑bit model and scored 82.47 per cent on the MRPC benchmark, outperforming DistilBERT despite using 55 times less training data. The paper will be presented at the GlobalSouthAI workshop at NeurIPS.
Disclaimer : This story is auto aggregated by a computer programme and has not been created or edited by DOWNTHENEWS. Publisher: deccanchronicle.com
