thinker:openalex-A5063383301Can Rager
Authored papers (8)
Llama-3.1-8B solves cyclic arithmetic (e.g., "what month is six months after August?") not by performing modular addition in the period of the cyclic concept (12 for months, 7 for days of the week) as its circularly structured representations might suggest, but by reusing a single generic base-10 addition mechanism across all such tasks. The mechanism operates in two stages: first, the model computes the raw integer sum of its inputs (six + August = 14), then maps that sum back into cyclic concept space (14 → February). The paper introduces a combined causal-abstraction and feature-geometry analysis to demonstrate that Llama-3.1-8B uses task-agnostic Fourier features with periods characteristic of base-10 arithmetic—specifically periods 2, 5, and 10—rather than the concept-native period. Strikingly, a sparse set of only 28 MLP neurons at layer 18 (approximately 0.2% of that layer's MLP) is responsible for this computation across all cyclic tasks and can be partitioned into disjoint clusters, each computing the sum for a Fourier feature of a distinct period. This implies that structured representational geometry does not entail correspondingly structured computation: a model can build domain-specific circular embeddings and yet delegate the actual arithmetic to a domain-general, base-10 calculator reused opportunistically, suggesting that mechanistic interpretability analyses that read off algorithms from representation geometry alone will systematically mislead.
Sparse autoencoders trained on Llama3.1-8B (layer 19, 500M Pile tokens, 25 SAEs across Standard, JumpReLU, TopK, BatchTopK and Matryoshka architectures at VE > 0.85) do not capture continuous concept manifolds as compact subspaces; they spread them across support sets well beyond the manifold's ambient dimension, a fragmented regime the paper calls dilution. The theory separates two ways an SAE can capture a manifold: globally, with a small atom group whose span contains it (Theorem 1, requiring μ < 1/(2k−1) incoherence), or locally, by tiling it with localized features. Tiling splits into shattering (mutually exclusive atoms) and dilution (redundant, mixed-selectivity atoms). A synthetic benchmark of 48 manifold instances (8 types, embedded in R^128, dictionary c=512) shows a sparsity sweet spot near k=4 with shattering below it and dilution above. On Llama, restricted R² plateaus past the ambient dimension and features show tuning curves, for example ones-digit and decade features on the years manifold. The instrument is an Ising-coupling pipeline: a pairwise Ising model fit on binarized codes, then Louvain communities and a PCA spectral-gap check. It recovers temperature, colors and political bias and surfaces a new epistemic-uncertainty manifold, while decoder cosine and Pearson correlation fail. Low reconstruction error therefore does not mean coherent geometry, and single-feature interpretation, steering and auto-interp miss it. The paper argues the unit of interpretability should be geometric objects, meaning groups of atoms, and that featurizers should target manifolds directly.
Resampling noise in reasoning-chain uncertainty dynamics is almost exactly multinomial, so most of the cost of Forking Paths Analysis re-measures structure that neighboring tokens already contain. On tinyMMLU (n=100) with Llama-3-8B-Instruct and DeepSeek-R1-Distill-Llama-8B, 1.77B tokens of S=200 reference data show log-TVD falling with log S at slopes of −0.4903 (development set), −0.4757 (Llama) and −0.4838 (DeepSeek), against the −1/2 expected under i.i.d. multinomial sampling. Measured replicate TVD sits within 0.98–1.01 of the exact multinomial null. The paper's estimator, a three-stage smoother, exploits this: PELT change-point detection with an exact multinomial cost segments the base path at forking points, Gaussian-kernel-weighted Dirichlet pooling smooths within segments, and cross-validation tunes the penalty and bandwidth. Smoothing S=5 data multiplies effective sample size by 5.0× at 4-token spacing for Llama (22.1× at 1-token spacing) and 4.9× for DeepSeek sentence resampling, falling to about 1× at S=100. Combined with coarser resampling, it cuts total budget to 1/8 with small error increase, roughly 1% of reference cost in the Figure 1 example. Gains shrink at large forks: at a 0.20 threshold smoothing is marginally worse than raw data. The implication is that resampling-based analysis of reasoning should be treated as a statistical estimation problem with a known noise model, not a brute-force sampling problem, and that sampling budgets can be allocated accordingly.
- Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior2026merged
Manifold steering — intervening along activation-space paths defined by the intrinsic geometry of a fitted activation manifold M_h rather than along Euclidean linear directions — produces behavioral trajectories that remain on the behavior manifold M_y, whereas standard linear steering cuts through off-manifold regions and generates unnatural outputs. Across language model reasoning tasks with cyclic, sequential, and graph geometries, and in a video world model whose task geometry corresponds to physical dynamics, the bidirectional relationship M_h ↔ M_y holds: steering that respects M_h tracks M_y, and optimizing interventions in activation space to follow M_y recovers activation trajectories that trace the curvature of M_h. The method fits two manifolds — one to intermediate representations, one to output probability distributions — and then uses geodesic-like paths on M_h as the intervention substrate rather than a single steering vector. This work argues that neural representation geometry is not incidental but is the causal structure organizing model behavior, which implies that the core problem of model steering should be reframed from finding the right direction to finding the right geometry.
Manifold steering — intervening on model activations along paths constrained to lie on a learned activation manifold M_h rather than along Euclidean linear directions — produces behavioral trajectories that track the corresponding behavior manifold M_y, while linear (Euclidean) steering cuts through off-manifold regions and generates unnatural outputs. The paper fits M_h to internal representations and M_y to output probability distributions, then tests their bidirectional correspondence via controlled interventions across language models and a video world model. In language models, tasks with cyclic geometries, sequential geometries, and complex graph geometries (in-context learning) all show that manifold-constrained interventions keep behavioral outputs on M_y, whereas linear steering deviates measurably. In a video world model, interventions shaped by physical-dynamics geometry similarly respect M_y. Crucially, the relationship is bidirectional: optimizing interventions in activation space to produce paths along M_y recovers activation trajectories that trace the curvature of M_h. This implies that the core problem of mechanistic steering should be recast not as finding the right direction in a flat Euclidean activation space, but as identifying the correct geometric structure — because representational geometry is not incidental to model behavior but is causally constitutive of it.
Lifting the featurizer's atom from a direction to a block of directions recovers concept manifolds that SAEs shatter, and block sparsity is the MAP-matched prior for a representation that is a sparse Minkowski sum of low-dimensional manifolds. The paper introduces block-sparse featurizers (BSFs) in three variants: Vanilla (top-k blocks by norm, untied encoder), Grassmannian (tied encoder with orthonormal charts), and Group Lasso (block soft-threshold). On a synthetic manifold-superposition toy, all three reach per-block R² of 0.93–0.97 against an oracle of 0.99, while a TopK SAE collapses to 0.53. On DINOv3 ViT-B, a minimum-description-length analysis at the 20% distortion floor shows blocks of dimension 2–4 describe activations in fewer bits than b=1 SAEs at every dictionary width and sparsity, and stable rank saturates near 3 even when b=16 is allowed. In InceptionV1, the curve-detector neurons and SAE atoms are fragments of one curve manifold, whose first three Fourier harmonics carry 89% of variance (59/18/12%) and survive a matched null by factors of 470 and 314. A DINOv3 block tracks sun luminance and shadow across twelve Blender models, and SDXL blocks support manifold steering, as in the pretzel block. The implication is that a featurizer is a hypothesis about the data-generating process, so interpretability should infer what the atoms must be rather than assume them. Blocks are one answer, matched to vision but probably not to language.
Gains on the sparsity-fidelity frontier do not predict how well a sparse autoencoder performs on downstream tasks, and SAEBench demonstrates this by scoring over 200 SAEs on eight metrics. Matryoshka BatchTopK SAEs rank below TopK and BatchTopK on loss recovered, yet they take the best score on Absorption, RAVEL, Sparse Probing, SCR and TPP, which is 5 of 8 metrics, in the typical L0 range of 40-200. Matryoshka is also the only architecture whose disentanglement scores improve when width grows from 4k to 16k to 65k latents on Gemma-2-2B (layer 12). The other architectures show inverse scaling on absorption and SCR, and the effect persists when the number of ablated latents is held fixed. SAEBench combines four capability groups (concept detection, interpretability, reconstruction, feature disentanglement). It adds two novel zero-ablation metrics, Spurious Correlation Removal and Targeted Probe Perturbation, and it extends the absorption metric to partial and shared absorption. The suite covers seven architectures (ReLU, TopK, BatchTopK, Gated, JumpReLU, P-annealing, Matryoshka), Gemma-2-2B and Pythia-160M, and Gemma-Scope SAEs up to 1M latents. Hierarchical architectures beat the rest by 30-40% on SCR and absorption. No single L0 wins everywhere, though L0 of 50-150 is a reasonable compromise. The implication is that SAE progress needs a multi-metric benchmark, and that optimizing reconstruction at fixed sparsity can hide failures in concept isolation, which grow with scale. The paper declines to combine the metrics into one score.
More papers — OpenAlex / S2
Affiliations (1)
- Goodfire(institute)
Co-authors (12)
- Atticus Geiger47 shared
- Ekdeep Singh Lubana47 shared
- Jack Merullo47 shared
- Tal Haklay45 shared
- Thomas McGrath39 shared
- Thomas Fel38 shared
- Usha Bhalla38 shared
- Daniel Wurgaft36 shared
- Owen Lewis32 shared
- Sheridan Feucht32 shared
- Matthew Kowal30 shared
- Raphaël Sarfati23 shared
Their work is cited by (2)
Other inbound relations (3)
Recent mentions (3)
- papers-typed
feucht-goodfire-geometric-calculator-2026.md - papers-typedpaper.md
- papers-typedwurgaft-2026-manifold.md