NEW BIML Bibliography entry
https://arxiv.org/pdf/2603.28052
Meta-Harness: End-to-End Optimization of Model Harnesses
Lee, Yoonho, Roshen Nair, Qizheng Zhang, Kangwook Lee, Omar Khattab, and Chelsea Finn
Harnesses for Agentic AI include perception and memory devices that allow an LLM to externalize and preserve state. This work describes iterating over a set of harnesses and finding better ones. Results are impressive.
NEW BIML Bibliography entry
https://arxiv.org/abs/2512.24601
Recursive Language Models
Alex L. Zhang, Tim Kraska, Omar Khattab
An excellent paper describing how to extend prompt context with recursion. Simple experiments. Clear explanations. This one makes you think. Are we moving towards actual Hofstaderian strange loops?
NEW BIML Bibliography entry
https://arxiv.org/abs/2602.06923v1#
From Kepler to Newton: Inductive Biases Guide Learned World Models in Transformers
Ziming Liu, Sophia Sanborn, Surya Ganguli, Andreas Tolias
Representation matters and is deeply constrained by tokenization. Excellent work, clearly described with real substance.
NEW BIML Bibliography entry
https://arxiv.org/abs/2503.03150
Position: Model Collapse Does Not Mean What You Think
Rylan Schaeffer, Joshua Kazdan, Alvan Caleb Arulandu, Sanmi Koyejo
We think recursive pollution is a better term than model collapse. Weak terminology leads to misunderstanding of impact. See figure 4. This is a very good paper.
NEW BIML Bibliography entry
https://arxiv.org/abs/2410.04840
Strong Model Collapse
Elvis Dohmatob, Yunzhen Feng, Arjun Subramonian, Julia Kempe
(NYU and META)
Recursive pollution leads to model collapse. This view of strong model collapse describes what happens in the case of recursive data poison.
#TOPPAPER #MLsec #Data #RecursivePollution
NEW BIML Bibliography entry
https://arxiv.org/abs/2509.16499
A Closer Look at Model Collapse: From a Generalization-to-Memorization Perspective
Lianghe Shi, et al
A very nice set of references to work in model collapse. Collapsed model == lookup table (that is, no generalization). Discussion of recursive pollution as causing variance shrinkage or distribution shift.
NEW BIML Bibliography entry AND NEW TOP FIVE #MLsec PAPER
READ IT
https://arxiv.org/pdf/2510.07192
Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
Alexandra Souly, ... Nicholas Carlini, et al
Excellent paper, clear and well-stated (like all Carlini papers). This result shows that recursive pollution risk is even greater than we thought. Injecting backdoors is pretty easy. The examples are a bit simplistic.
NEW BIML Bibliography entry
https://arxiv.org/abs/2510.04871
Less is More: Recursive Reasoning with Tiny Networks
Alexia Jolicoeur-Martineau
This is an engineering exercise akin to “set it to 57,” but it is really interesting. A set of weekend kludges that has important implications. Harold wants to pursue this line to think about how integers are represented. Won the ARC prize.
NEW BIML Bibliography entry
https://www.nature.com/articles/s41586-025-09446-5
Optical generative models
Chen, Shiqi, Yuhang Li, Yuntian Wang, Hanlong Chen and Aydogan Ozcan
Light for computation with properties of low power and superposition. Analo of quantum computing. This reminds os of Rosenblatt’s Perceptrons from the ’50s.
NEW BIML Bibliography entry
https://direct.mit.edu/books/oa-monograph/5600/Context-Changes-EverythingHow-Constraints-Create
Chapter 13, Context Changes Everything
Alicia Juarrero
A solid treatment of the 4Es theory (Embodied, Embedded, Extended, Enactive) properly grounded in philosophy of mind.
Gary McGraw