Each Sunday, NPR host Will Shortz, The New York Instances’ crossword puzzle guru, will get to quiz 1000’s of listeners in a long-running phase known as the Sunday Puzzle. Whereas written to be solvable with out too a lot foreknowledge, the brainteasers are often difficult even for expert contestants.
That’s why some consultants suppose they’re a promising option to check the bounds of AI’s problem-solving skills.
In a latest research, a workforce of researchers hailing from Wellesley Faculty, Oberlin Faculty, the College of Texas at Austin, Northeastern College, Charles College, and startup Cursor created an AI benchmark utilizing riddles from Sunday Puzzle episodes. The workforce says their check uncovered stunning insights, like that reasoning fashions — OpenAI’s o1, amongst others — typically “hand over” and supply solutions they know aren’t right.
“We wished to develop a benchmark with issues that people can perceive with solely basic information,” Arjun Guha, a pc science school member at Northeastern and one of many co-authors on the research, informed TechCrunch.
The AI trade is in a little bit of a benchmarking quandary in the meanwhile. Many of the assessments generally used to judge AI fashions probe for expertise, like competency on PhD-level math and science questions, that aren’t related to the common person. In the meantime, many benchmarks — even benchmarks launched comparatively just lately — are shortly approaching the saturation level.
The benefits of a public radio quiz recreation just like the Sunday Puzzle is that it doesn’t check for esoteric information, and the challenges are phrased such that fashions can’t draw on “rote reminiscence” to unravel them, defined Guha.
“I believe what makes these issues arduous is that it’s actually troublesome to make significant progress on an issue till you clear up it — that’s when all the things clicks collectively ,” Guha mentioned. “That requires a mixture of perception and a strategy of elimination.”
No benchmark is ideal, in fact. The Sunday Puzzle is U.S. centric and English solely. And since the quizzes are publicly obtainable, it’s potential that fashions skilled on them can “cheat” in a way, though Guha says he hasn’t seen proof of this.
“New questions are launched each week, and we are able to anticipate the most recent inquiries to be actually unseen,” he added. “We intend to maintain the benchmark recent and observe how mannequin efficiency modifications over time.”
On the researchers’ benchmark, which consists of round 600 Sunday Puzzle riddles, reasoning fashions corresponding to o1 and DeepSeek’s R1 far outperform the remaining. Reasoning fashions totally fact-check themselves earlier than giving out outcomes, which helps them keep away from a number of the pitfalls that usually journey up AI fashions. The trade-off is that reasoning fashions take a little bit longer to reach at options — usually seconds to minutes longer.
At the very least one mannequin, DeepSeek’s R1, offers options it is aware of to be unsuitable for a number of the Sunday Puzzle questions. R1 will state verbatim “I hand over,” adopted by an incorrect reply chosen seemingly at random — conduct this human can actually relate to.
The fashions make different weird decisions, like giving a unsuitable reply solely to right away retract it, try and tease out a greater one, and fail once more. In addition they get caught “considering” without end and provides nonsensical explanations for solutions, or they arrive at an accurate reply instantly however then go on to contemplate different solutions for no apparent cause.
“On arduous issues, R1 actually says that it’s getting ‘pissed off,’” Guha mentioned. “It was humorous to see how a mannequin emulates what a human may say. It stays to be seen how ‘frustration’ in reasoning can have an effect on the standard of mannequin outcomes.”
The present best-performing mannequin on the benchmark is o1 with a rating of 59%, adopted by the just lately launched o3-mini set to excessive “reasoning effort” (47%). (R1 scored 35%.) As a subsequent step, the researchers plan to broaden their testing to further reasoning fashions, which they hope will assist to determine areas the place these fashions could be enhanced.
“You don’t want a PhD to be good at reasoning, so it must be potential to design reasoning benchmarks that don’t require PhD-level information,” Guha mentioned. “A benchmark with broader entry permits a wider set of researchers to understand and analyze the outcomes, which can in flip result in higher options sooner or later. Moreover, as state-of-the-art fashions are more and more deployed in settings that have an effect on everybody, we consider everybody ought to be capable of intuit what these fashions are — and aren’t — able to.”
