PhD Students Weida Li and Zhuanghua Liu Win Best Paper Runner-up Award at UAI 2026

Two of our Computer Science PhD students, Weida Li and Zhuanghua Liu, have won the Best Paper - Runner-up Award at the Conference on Uncertainty in Artificial Intelligence (UAI) 2026 — ranking 2nd among 1,087 submissions to one of the top-ranked venues in AI research, working under the supervision of Associate Professor Bryan Kian Hsiang Low.

Their paper, "Nonlinear Axiomatic Attribution for Cooperative Games," tackles a quiet flaw in one of machine learning's most trusted tools: the Shapley value, widely used to rank how much each player, feature, or data point contributes to an outcome. The team found that its defining strength, linearity, is also its blind spot – it can miss real differences between contributors because too much variation falls into what the method simply can't see. Their fix borrows from the "least core," a longstanding alternative to Shapley values, and builds a new family of nonlinear attribution methods around it, each one solving an optimisation problem to stay as faithful as possible to the real data.

Attribution methods like this quietly shape how AI decisions get explained, how contributions get priced in data markets, and how audits trace which inputs actually mattered.

Congratulations to Weida, Zhuanghua, and Prof Low!