Overfitting
8
MENTIONS
7
EPISODES
7
PODCASTS
Search complete. 8 mentions across 7 episodes found for "Overfitting".
Sep 9, 2026
How to Generate Synthetic Pretraining Data? with Joël Niklaus from Hugginface
J
66:05Joel NiklausGUEST
But yeah, I, I don't know.
J
66:06Joel NiklausGUEST
To me, it seems a bit over-engineered and a bit, uh, like overfitted [chuckles] in the end.
J
66:12Joel NiklausGUEST
So yeah, I didn't do that.
J
66:14Joel NiklausGUEST
So yeah, it, it looks like, um, there really isn't a, a good proxy that you can use, and you really need to train and evaluate each data set independently, it looks like, unfortunately.
Mr. Nuge
N
38:31Noah BloombergHOST
detent.
N
38:32NugeGUEST
Like we're still going to be, we're getting all the parts overfit.
N
38:35NugeGUEST
Like the lock face is going to be, uh, what's called where I wouldn't say overfit.
N
38:39NugeGUEST
I mean, oversized.
N
38:41NugeGUEST
So I'm going to be having to grind the lock faces and then tune everything.
Data Centers are Good Actually, with Stephen Balaban
J
22:15JaydenHOST
If you overfit enough, an LLM is going to look amazing.
J
22:18JaydenHOST
Like as if you overfit anything.
J
22:21JaydenHOST
Overfit literally anything.
J
22:23JaydenHOST
You backtest anything in a particular timeframe and you can put your sharp ratio through the roof.
J
22:29JaydenHOST
But with that sort of thing, I always find it hard to believe that those that are sort of like, okay, well cool, we got LLMs trading our portfolio, it's like to go up against Jane Street, like what do you have that Jane Street does not have? What do you have that Citadel does not have? Like this is like the kind of engineering that a Jane Street will have.
P
Unknown podcast
The Quantum AI Digest: 19 August 2026 Daily Dispatch
Aug 19 · 1 Mention
S
9:45speaker_0NARRATOR
Studies reveal that safety alignment in open-weight large language models can be easily bypassed, and so-called decoy hardening is proving unreliable.
S
9:55speaker_0NARRATOR
Benchmarkpocalypse This is growing worry, as LLM agents are increasingly found to overfit or game performance benchmarks, leading to exaggerated claims that do not reflect real-world capabilities.
S
10:08speaker_0NARRATOR
Workforce trends show that while AI is generating new job categories, games are often concentrated among larger firms.
S
10:16speaker_0NARRATOR
Scholars warned that over-automation at the entry level could jeopardize the future talent pipeline, raising the risk of a tragedy of the commons for professional oversight of AI systems.
JJ Zachariason's Fantasy Football Draft Strategies (Re-Release)
J
36:15JJ ZachariasonGUEST
And so I do things like adding a, uh, you know, a, a program like a strength of schedule factor and, uh, you know, and, and age and a breakout factor and stuff like that.
J
36:25JJ ZachariasonGUEST
So I'm testing all of these collegiate metrics, and then eventually I kinda bring them back down to only having a handful because very easy to overfit with modeling.
J
36:35JJ ZachariasonGUEST
Overfit meaning you're taking Puka Nacua's rookie season and just associating all of his traits, assuming that future wide receivers should be more like Puka Nacua when the model should be realizing he's just more of an outlier.
A
36:50Alfredo BrownHOST
Right.
J
36:50JJ ZachariasonGUEST
Right?
P
Unknown podcast
The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future
Aug 16 · 1 Mention
S
14:22speaker_0HOST
Right.
S
14:23speaker_0HOST
Overfitting is a concept from machine learning, and it happens when a model learns the training data too well, to the point that it can't handle any new unexpected information.
S
14:31speaker_1HOST
Yeah, it gets stuck.
S
14:32speaker_0HOST
It's like having a personal chef who figures out I love pizza.
110. Calibrating Models: From Theory to Real‑World Accuracy
T
12:21Tina CarsonHOST
And understanding why reveals one of the deepest tensions in quantitative finance.
T
12:27Tina CarsonHOST
Overfitting occurs when a model captures noise rather than signal.
T
12:30Tina CarsonHOST
Think of a student who memorizes answers to specific test questions rather than learning the underlying concepts.
T
12:37Tina CarsonHOST
They might ace that particular test, but give them a slightly different set of questions and they fail.