Modern LLMs are increasingly deep, and depth correlates with performance, albeit with diminishing returns. However, do these models use their depth efficiently? Do they compose more features to create higher-order computations that are impossible in shallow models, or do they merely spread the same kinds of computation out over more layers? To address these questions, we analyze the residual stream of the Llama 3.1, Qwen 3, and OLMo 2 family of models. We find: First, comparing the output of the sublayers to the residual stream reveals that layers in the second half contribute much less than those in the first half, with a clear phase transition between the two halves.

Sparse autoencoders (SAEs) are designed to extract interpretable features from language models by enforcing a sparsity constraint. Ideally, training an SAE would yield latents that are both sparse and semantically meaningful. However, many SAE latents activate frequently (i.e., are dense), raising concerns that they may be undesirable artifacts of the training procedure. In this work, we systematically investigate the geometry, function, and origin of dense latents and show that they are not only persistent but often reflect meaningful model representations. We first demonstrate that dense latents tend to form antipodal pairs that reconstruct specific directions in the residual stream, and that ablating their subspace suppresses the emergence of new dense features in retrained SAEs--suggesting that high density features are an intrinsic property of the residual space. We then introduce a taxonomy of dense latents, identifying classes tied to position tracking, context binding, entropy regulation, letter-specific output signals, part-of-speech, and principal component reconstruction. Finally, we analyze how these features evolve across layers, revealing a shift from structural features in early layers, to semantic features in mid layers, and finally to output-oriented signals in the last layers of the model. Our findings indicate that dense latents serve functional roles in language model computation and should not be dismissed as training noise.

Retrieval-Augmented Generation (RAG) provides additional contextual knowledge to complement the parametric knowledge in Large Language Models (LLMs). These two knowledge interweave to enhance the accuracy and timeliness of LLM responses. However, the internal mechanisms by which LLMs utilize these knowledge remain unclear. We propose modeling the forward propagation of knowledge as an entity flow, employing this framework to trace LLMs' internal behaviors when processing mixed-source knowledge. Linear probing utilizes a trainable linear classifier to detect specific attributes in hidden layers.

Recent theoretical results show transformers cannot express sequential reasoning problems over long inputs, intuitively because their computational depth is bounded. However, prior work treats the depth as a constant, leaving it unclear to what degree bounded depth may suffice for solving problems over short inputs, or how increasing the transformer's depth affects its expressive power. We address these questions by analyzing transformers whose depth can grow minimally with context length n. We show even highly uniform transformers with depth Θ(logn) can express two important problems: recognizing regular languages, which captures state tracking abilities and was known to be expressible only by an unconventional, non-uniform model of transformers, and graph connectivity, which underlies multistep reasoning. Notably, both of these problems cannot be expressed by fixed-depth transformers under standard complexity conjectures, demonstrating the expressivity benefit of growing depth. Moreover, our theory quantitatively predicts how depth must grow with input length to express these problems, showing that depth scaling is more efficient than scaling width or chain-of-thought steps. Empirically, our detailed experiments designed to bridge the expressivity vs. learnability gap reveal that our theoretical depth requirements for regular language recognition closely match the practical depth requirements for successfully training transformers. Thus, our results clarify how depth affects a transformer's reasoning capabilities, and provide practical guidance for effective depth selection for sequential reasoning.

Large language models (LLMs) often concentrate their attention on a few specific tokens referred to as attention sinks. Common examples include the first token, a prompt-independent sink, and punctuation tokens, which are prompt-dependent. While the tokens causing the sinks often lack direct semantic meaning, the presence of the sinks is critical for model performance, particularly under model compression and KV-caching. Despite their ubiquity, the function, semantic role, and origin of attention sinks--especially those beyond the first token--remain poorly understood. In this work, we conduct a comprehensive investigation demonstrating that attention sinks: catch a sequence of tokens, tag them using a common direction in embedding space, and release them back into the residual stream, where tokens are later retrieved based on the tags they have acquired. Probing experiments reveal these tags carry semantically meaningful information, such as the truth of a statement. These findings extend to reasoning models, where the mechanism spans more heads and explains greater variance in embeddings, or recent models with querykey normalization, where sinks remain just as prevalent. To encourage future theoretical analysis, we introduce a minimal problem which can be solved through the'catch, tag, release' mechanism, and where it emerges through training.

What computational structure are we building into large language models when we train them on next-token prediction? Here, we present evidence that this structure is given by the meta-dynamics of belief updating over hidden states of the data-generating process. Leveraging the theory of optimal prediction, we anticipate and then find that belief states are linearly represented in the residual stream of transformers, even in cases where the predicted belief state geometry has highly nontrivial fractal structure. We investigate cases where the belief state geometry is represented in the final residual stream or distributed across the residual streams of multiple layers, providing a framework to explain these observations. Furthermore we demonstrate that the inferred belief states contain information about the entire future, beyond the local next-token prediction that the transformers are explicitly trained on. Our work provides a general framework connecting the structure of training data to the geometric structure of activations inside transformers.

In Section 3, we verify that this geometry is linearly represented in the residual stream of transformers.

We show how to "compile" human-readable programs into standard decoder-only transformer models.