English
Related papers

Related papers: Directional Optimization Asymmetry in Transformers…

200 papers

Despite the numerous advances, reinforcement learning remains away from widespread acceptance for autonomous controller design as compared to classical methods due to lack of ability to effectively tackle the reality gap. The reliance on…

Machine Learning · Computer Science 2024-09-23 Narendra Patwardhan , Zequn Wang

As we show in this paper, the prediction for output token $n+1$ of Transformer architectures without one of the mechanisms of positional encodings and causal attention is invariant to permutations of input tokens $1, 2, ..., n-1$. Usually,…

Machine Learning · Computer Science 2024-06-18 Chunsheng Zuo , Michael Guerzhoy

Transformers have supplanted recurrent models in a large number of NLP tasks. However, the differences in their abilities to model different syntactic properties remain largely unknown. Past works suggest that LSTMs generalize very well on…

Computation and Language · Computer Science 2020-10-09 Satwik Bhattamishra , Kabir Ahuja , Navin Goyal

Transformers trained in low precision can suffer forward-error amplification. We give a first-order, module-wise theory that predicts when and where errors grow. For self-attention we derive a per-layer bound that factorizes into three…

Machine Learning · Computer Science 2025-10-28 Jinwoo Baek

Cross-entropy pretraining and preference alignment update the same transformer weights, but leave geometrically distinct traces. We characterise this asymmetry with a relative-subspace-fraction probe that tracks how weight deltas align with…

Machine Learning · Computer Science 2026-05-19 Valeria Ruscio , Eli-Shaoul Khedouri , Keiran Thompson

The determination of confidence intervals of stress orientation is a crucial element in the discussion of homogeneity or heterogeneity of the stress field under study. The error estimates provided by the grid search method Focal Mechanism…

Geophysics · Physics 2010-08-04 Stefan A. Revets

Alignment is vital for safely deploying large language models (LLMs). Existing techniques are either reward-based (training a reward model on preference pairs and optimizing with reinforcement learning) or reward-free (directly fine-tuning…

Computation and Language · Computer Science 2026-03-03 Ruoxi Cheng , Haoxuan Ma , Weixin Wang , Ranjie Duan , Jiexi Liu , Xiaoshuang Jia , Simeng Qin , Xiaochun Cao , Yang Liu , Xiaojun Jia

We study learned memory tokens as a computational scratchpad for a single-block Universal Transformer with Adaptive Computation Time (ACT) on Sudoku-Extreme, a combinatorial reasoning benchmark. Memory tokens are empirically necessary: no…

Machine Learning · Computer Science 2026-05-05 Grigory Sapunov

Graph Transformers can mix information globally, but this flexibility also creates failure modes: some tasks require long-range communication while others are better served by local interaction. We study this through a synthetic…

Machine Learning · Computer Science 2026-04-27 Qinhan Hou , Jing Tang

Language models are increasingly capable, yet still fail at a seemingly simple task of multi-digit multiplication. In this work, we study why, by reverse-engineering a model that successfully learns multiplication via \emph{implicit…

Machine Learning · Computer Science 2025-10-02 Xiaoyan Bai , Itamar Pres , Yuntian Deng , Chenhao Tan , Stuart Shieber , Fernanda Viégas , Martin Wattenberg , Andrew Lee

Do reasoning models have "Aha!" moments? Prior work suggests that models like DeepSeek-R1-Zero undergo sudden mid-trace realizations that lead to accurate outputs, implying an intrinsic capacity for self-correction. Yet, it remains unclear…

Artificial Intelligence · Computer Science 2026-04-21 Liv G. d'Aliberti , Manoel Horta Ribeiro

The reversal curse describes a failure of autoregressive language models to retrieve a fact in reverse order (e.g., training on ``$A > B$'' but failing on ``$B < A$''). Recent work shows that objectives with bidirectional supervision (e.g.,…

Computation and Language · Computer Science 2026-04-08 Julian Coda-Forno , Jane X. Wang , Arslan Chaudhry

In modern LLMs, linguistic features function not as stylistic artifacts but as probes of probability mass, allocated under training alignment objectives. Language models trained with contemporary pipelines exhibit severe reshaping of…

Computation and Language · Computer Science 2026-05-29 Rohan Mahapatra

Autoregressive language models (ARMs) suffer from the reversal curse: after learning ''$A$ is $B$,'' they often fail on the reverse query ''$B$ is $A$.'' Masked diffusion language models (MDMs) exhibit this failure in a much weaker form,…

Artificial Intelligence · Computer Science 2026-05-13 Moongyu Jeon , Sangwoo Shin , BumJun Kim , Kyelim Lee , Albert No

Many learning problems require uncovering a hidden ordering that reveals structure in unordered data, such as monotonicity in sorting or spatial continuity in jigsaw reconstruction. In these settings, permutations can be learned as latent…

Machine Learning · Computer Science 2026-05-26 Ran Eisenberg , Ofir Lindenbaum

Transformers have demonstrated exceptional performance across a wide range of domains. While their ability to perform reinforcement learning in-context has been established both theoretically and empirically, their behavior in…

Machine Learning · Statistics 2025-10-24 Baiyuan Chen , Shinji Ito , Masaaki Imaizumi

Large Language Models (LLMs) achieve strong performance on diverse tasks but often exhibit cognitive inertia, struggling to follow instructions that conflict with the standardized patterns learned during supervised fine-tuning (SFT). To…

Grokking -- the delayed transition from memorization to generalization in small algorithmic tasks -- remains poorly understood. We present a geometric analysis of optimization dynamics in transformers trained on modular arithmetic. PCA of…

Machine Learning · Computer Science 2026-04-06 Yongzhong Xu

This paper explores continuous-time control synthesis for target-driven navigation to satisfy complex high-level tasks expressed as linear temporal logic (LTL). We propose a model-free framework using deep reinforcement learning (DRL) where…

Robotics · Computer Science 2023-03-17 Mingyu Cai , Makai Mann , Zachary Serlin , Kevin Leahy , Cristian-Ioan Vasile

Large Language Models (LLMs) are typically trained to predict in the forward direction of time. However, recent works have shown that prompting these models to look back and critique their own generations can produce useful feedback.…

Computation and Language · Computer Science 2025-02-04 Yerram Varun , Rahul Madhavan , Sravanti Addepalli , Arun Suggala , Karthikeyan Shanmugam , Prateek Jain