When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation
55 points - today at 4:10 PM
SourceComments
If the Pareto rule is any indicator, 80% of results come from 20% of causes. It seems that we have alot more to learn about intelligence.
I am reminded of a statement on truth from an ancient philosopher, this sentiment seems to be exactly the opposite of the LLM training paradigm
āThe seeker after truth is not one who studies the writings of the ancients and, following his natural disposition, puts his trust in them, but rather the one who suspects his faith in them and questions what he gathers from them, the one who submits to argument and demonstration and not the sayings of human beings whose nature is fraught with all kinds of imperfection and deficiency. Thus the duty of the man who investigates the writings of scientists, if learning the truth is his goal, is to make himself an enemy of all that he reads, and, applying his mind to the core and margins of of its content, attack it from every side. he should also suspect himself as he performs his critical examination of it, so that he may avoid falling into either prejudice or leniency.ā - ibn al-Haytham
Data at https://gertlabs.com/rankings
Oh! I got my name on a paper! I don't think there is much reward for it these days.
Benchmarks are handy when theyāre new, novel, and constantly changing. The second you let even a single aspect of it stagnate, it becomes a gameable score rather than a useful metric. In PC Gaming, we saw vendors optimize for specific titles, benchmark tools, and scenarios at the expense of general performance, and eventually the industry had a ācome to Jesusā moment where we had to collectively decide how to move forward from an industry built on thoroughly gamed benchmarks, with entities like Gamersā Nexus and Digital Foundry being the end results of that falling out.
LLMs were always going to end up the same way, because the people building the benchmarks - well-intentioned as they were - ultimately fell into the exact same traps with fixed scoring rubrics, known test questions, and believing in some form of ācompletenessā that could be attained or achieved. The net result are models consistently scoring better on benchmarks but also seeing diminishing returns and rising vulnerabilities, because actual improvement or utility isnāt what theyāre being optimized for so much as bragging rights. Itās why thereās so much growing interest in things like MoE execution on unified memory platforms as a means of porting larger models to consumer kit, or ternary models (shoutout to Bonsai) as a means of reducing overall size: both take leading edge, benchmark-saturating models and show that with minimal score loss, they function about as well as frontier models might.
Building a new benchmark wonāt solve the problem, either. To move forward, we must evaluate LLMs objectively and with continuously evolving workloads. More āpelican on a bicycleā stuff, but from varying perspectives and use cases. Radiologists putting models through their paces with usable sample data they donāt share with AI labs, or IT folks tasking agents with bootstrapping specific, real-world workloads. To prove general intelligence, we need more specialists evaluating them specifically and generally in ways that are transparent to consumers but difficult or impossible for AI companies to prepare against.
Only then will scoring values matter.
This results in a lot of "oh wow it can do math I dont care about" and "it can't code a lot, but not well" outcomes instead of the core needs:
1) Cheaper faster and real time 2) Long walk capable without losing attention while rescoring goals over updated enviroment 3) Specific domain knowledge that can be trained quickly into the model (how we do work in this specific case)
> We find that nearly half of the our bench- marks exhibit saturation