Singapore

Recent studies have revealed that LLMs can exhibit behavioral self-awareness — the ability to accurately describe or predict their own learned behaviors without explicit supervision. This capability raises safety concerns, as it may allow models to intentionally conceal their true abilities during evaluation. We attempt to better understand this phenomenon by investigating how and when self-awareness emerges during fine-tuning, and whether it can be mechanistically localized. Through controlled fine-tuning experiments on instruction-tuned LLMs with low-rank adapters (LoRA), we find: (1) that across domains and layers, fine-tuning consistently elicits self-aware behavior early on in the training process; (2) that self-awareness can be reliably induced using a single rank-1 LoRA adapter; and (3) that the learned self-aware behavior is largely captured by a single steering vector in activation space, recovering nearly all of the fine-tune’s behavioral effect. Together, these findings suggest that self-awareness exhibits a rapid transition and is captured by a linear direction rather than a distributed introspective mechanism.

AAAI 2026

Emergent Mechanisms of Self-Awareness in LLMs

nlp: safety and robustness

and evaluation of nlp models secondary: nlp: (large) language models

primary: nlp: interpretability

explainable ml

ml: transparent

interpretable

analysis

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
 The conference will be held in Toronto, Canada, July 9-14, 2023. 
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience. 
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**

ACL 2023

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

workshop paper

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

We introduce a diagnostic for symmetry learning in PDE surrogates: the influence function computed across symmetry- related states. On compressible Euler flows, our diagnostic reveals that a UNet exhibits partial but unstable influence across square group actions and translations, whereas a ViT reaches lower prediction error yet shows largely orthogonal updates across orbits. This exposes an optimization-symmetry tradeoff: stronger inductive biases promote data efficiency but can couple updates rigidly; flexible architectures optimize easily but ignore physical structure. Our diagnostic offers a reproducible test for whether training dynamics propagate information across symmetry orbits, a necessary ingredient for robust generalization in scientific machine learning.

Loss Landscape Geometry and the Learning of Symmetries: Or, What Influence Functions Reveal About Memorization and Generalization

Additive manufacturing enables the layer-by-layer fabrication of complex and customizable structures but remains prone to internal defects that are difficult to detect during fabrication. This work presents an anomaly detection framework for real-time defect monitoring using a custom-designed wireless antenna sensor. The sensor consists of an aluminum split-ring resonator patterned with gold conductive ink and mounted on a ceramic substrate. It passively responds to structural and thermal changes during the manufacturing process by producing frequency-domain scattering parameter signals. To process these high-dimensional signals without requiring defect-labeled data, we develop a zero-bias deep neural network trained exclusively on normal-class samples. The model constructs a class-structured latent space using cosine similarity to learned templates and applies Mahalanobis distance to define statistical decision boundaries. Anomalies are identified when inputs deviate significantly from the expected distribution of normal data. The framework is designed around two deployment-oriented principles. Explainability enables users and domain experts to understand detection decisions through latent space visualization, feature attribution, and statistical reasoning that connects model outputs to interpretable signal characteristics. Instructibility allows users to tune detection sensitivity through threshold control without retraining, adapting the system to different quality requirements and process tolerances. The system is evaluated using experimental data collected from the antenna sensor under controlled thermal variation. Results demonstrate that the proposed method accurately detects anomalous signal patterns without using defect supervision, while maintaining interpretability and user control. This approach offers a transparent, adaptive, and data-efficient solution for in situ defect monitoring in additive manufacturing environments.

Anomaly Detection for Antenna-Based Defect Detection in Additive Manufacturing: An Explainable, Instructive Learning Framework

Reports of “emergent” capabilities in transformer-based LLMs (abrupt, non-linear improvements in task performance) remain controversial due to post-hoc measurement and ambiguous definitions. We investigate whether such transitions can be predicted pre-hoc from internal representations. For each training checkpoint, we train sparse autoencoders (SAEs) on model activations, construct a co-activation graph over SAE features, and track graph statistics (e.g. density, clustering, etc.). We analyze a 2-layer grokking-style transformer with aggregate graph metrics over eight SAE initializations, and test lead-lag relationships between changes in graph metrics and subsequent changes in accuracy under a formalized emergence criterion. Across both settings, we find no statistically significant evidence that global co-activation topology forecasts emergent jumps in performance. If pre-hoc indicators exist, they may lie outside the global graph measures analyzed here (e.g., in task-specific circuits or localized subgraphs).

Predicting Emergent Capabilities Using Sparse Features

We present a differentiable framework that leverages the Discrete Empirical Interpolation Method (DEIM) for interpretable deep learning and dynamical system analysis. Although DEIM efficiently approximates nonlinear terms in projection-based reduced-order models (POD-ROM), its fixed interpolation points limit the adaptability to complex and time-varying dynamics. To address this limitation, we first develop a differentiable adaptive DEIM formulation for the one-dimensional viscous Burgers’ equation, which allows neural networks to dynamically select interpolation points in a computationally efficient and physically consistent manner. We then apply DEIM as an interpretable analysis tool for examining the learned dynamics of a pre-trained Neural Ordinary Differential Equation (NODE) on a two-dimensional vortex-merging problem. The DEIM trajectories reveal physically meaningful features in the NODE’s learned dynamics and expose its limitations when extrapolating to unseen flow configurations. These findings demonstrate that DEIM can serve not only as a model reduction tool but also as a diagnostic framework for understanding and improving the generalization behavior of neural differential equation models.

Towards Interpretable Deep Learning and Analysis of Dynamical Systems via the Discrete Empirical Interpolation Method

Classifying non-Hermitian energy spectra under open boundary conditions is an open challenge in physics. This classification is a critical prerequisite for the rational inverse design of systems exhibiting desired dynamics and topological responses. While graph topology has emerged as a promising approach for characterizing these spectra, systematic methods for distilling non-Hermitian spectra into their corresponding graph representations have been lacking. Moreover, the resulting graphs often exhibit complexities that defy manual classification, necessitating machine learning approaches. In this work, we introduce a two-step framework for classifying non-Hermitian spectra based on their graph topologies. The first step employs $\texttt{Poly2Graph}$, an automated, high-performance pipeline that distills non-Hermitian spectra into $\textit{spectral graphs}$ suitable for graph neural networks (GNNs). The second step involves generating a large dataset of these spectral graphs and training a GNN for classification. We propose $\texttt{GnLTransformer}$, a novel architecture featuring dual channels that leverage line graphs to explicitly capture higher-order topological features. $\texttt{GnLTransformer}$ achieves over 99% classification accuracy on our dataset, outperforming standard baselines by 32%. Notably, beyond conventional GNNs, $\texttt{GnLTransformer}$ offers inherent explainability regarding higher-order topology. As a further contribution, we release a new $\textit{multi-graph}$ dataset comprising over 117K spectral graphs.

Classifying the Graph Topology of Non-Hermitian Energy Spectra with Graph Transformer

Machine learning force fields (MLFFs) now rival quantum methods in accuracy, yet their internal logic remains opaque—a critical barrier to both trust and scientific discovery. We introduce **DUAL-X**, a *Dual-Level Explainability Framework* that bridges model reasoning with human understanding. **DUAL-X** unites two complementary perspectives: a **model-centric** level identifying *where* in the atomic structure the model focuses its attention, and a **human-centric** level revealing *what* physically meaningful interactions it prioritizes. Implemented with Grad-CAM for spatial attribution and SHAP-on-SOAP for physical interpretation, **DUAL-X** provides a general, human-in-the-loop paradigm for interpretable scientific AI. Applied to dopant migration—a challenging task for MLFFs—**DUAL-X** reveals that training strategy governs the emergence of chemical intuition. Multi-temperature fine-tuned models exhibit over $10^2\times$ stronger selectivity for Cr–Cr $f$-type ($l=3$) angular correlations than scratch-trained ones, emphasizing complex 3D coordination motifs essential for accuracy. Models with sharper Grad-CAM focus also display coherent SHAP importance for dopant clustering, revealing consistent internal reasoning across scales. By aligning model logic with human physical knowledge, **DUAL-X** transforms opaque predictors into interpretable scientific partners—advancing trustworthy, explainable, and insight-driven AI for materials discovery.

What is Your Force Field Really Learning? Gaining Scientific Intuition with A Dual-Level Explainability Framework

In data-scarce settings, learning accurate and interpretable models for high-stakes medical tabular classification remains a fundamental challenge, as healthcare decisions must be transparent and trustworthy. We propose a novel pipeline for few-shot tabular classification that distills the predictive power of large foundation models into globally interpretable student models. Given a tabular dataset, we generate synthetic data using CTGAN to approximate the underlying distribution. We then train high-capacity teacher models (TabPFN, TabM) on a small number of labeled examples and use them to pseudolabel the synthetic data. Finally, we train student explainer models (i.e., XGBoost, decision trees, Generalized Linear Rule Model (GLRM), and TTnet) on this pseudolabeled synthetic dataset. These student models are globally and exactly interpretable, yielding logical decision rules (e.g., disjunctive normal form) that fully reproduce their predictions. Evaluated across 8 clinical tabular tasks, our distilled models generally outperform baselines trained directly on the few-shot data, with improved ROC AUC scores across few-shot settings. This work demonstrates that foundation models can be effectively leveraged as teachers to produce small, transparent, and high-performing classifiers. Our approach advances the goal of reliable and interpretable machine learning in real-world settings where labeled data is limited.

Leveraging Foundation Models in Healthcare: A Distillation Approach to Interpretable Clinical Prediction

Adjoint-based inverse design yields compact, high-performance nanophotonic devices, but the mapping from pixel-level layouts to optical figures of merit remains hard to interpret. We present a simple pipeline that (i) generates a large set of wavelength demultiplexers (WDMs) with SPINS-B, (ii) records each final 2D layout and its spectral metrics (e.g., transmitted power at 1310 nm and 1550 nm), and (iii) trains a lightweight convolutional surrogate to predict these metrics from layouts, enabling (iv) gradient-based attribution via Integrated Gradients (IG) to highlight specific regions most responsible for performance. On a corpus of sampled WDMs, IG saliency consistently localizes to physically meaningful features (e.g., tapers and splitter hubs), offering design intuition that complements adjoint optimization. Our contribution is an end-to-end, data-driven workflow—SPINS-B dataset, CNN surrogate, and IG analysis—that turns inverse-designed layouts into interpretable attributions without modifying the physics solver or objective, and that can be reused for other photonic components.

Towards Explainable Inverse Design for Photonics via Integrated Gradients

Recent research has extensively studied how large language models manipulate integers in specific arithmetic tasks, and on a more fundamental level, how they represent numeric values. These previous works have found that language model embeddings can be used to reconstruct the original values, however, they do not evaluate whether language models actually model continuous values as continuous. Using expected properties of the embedding space, including linear reconstruction and principal component analysis, we show that language models not only represent numeric spaces as non-continuous but also introduce significant noise. Using models from three major providers (OpenAI, Google Gemini and Voyage AI), we show that while reconstruction is possible with high fidelity ($R^2 \geq 0.95$), principal components only explain a minor share of variation within the embedding space. This indicates that many components within the embedding space are orthogonal to the simple numeric input space. Further, both linear reconstruction and explained variance suffer with increasing decimal precision, despite the ordinal nature of the input space being fundamentally unchanged. The findings of this work therefore have implications for the many areas where embedding models are used, in-particular where high numerical precision, large magnitudes or mixed-sign values are common.

Downloads

Next from AAAI 2026

Loss Landscape Geometry and the Learning of Symmetries: Or, What Influence Functions Reveal About Memorization and Generalization

Similar lecture

Don't Lose Yourself! Empathetic Response Generation via Explicit Self-Other Awareness

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2026

Loss Landscape Geometry and the Learning of Symmetries: Or, What Influence Functions Reveal About Memorization and Generalization

Similar lecture

Don't Lose Yourself! Empathetic Response Generation via Explicit Self-Other Awareness

Downloads