- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2(+1)
Read review
Paper summary
The paper investigates how compositional reasoning emerges during learning through three main aspects: law-specific emergence, structural identifiability, and relation integration. The authors analyze the learning dynamics of shallow and deep linear neural networks and develop a theory linking the structure of training evidence to the emergence of compositional inference. Their findings explain why a model may fail to perform compositional inference despite learning the premises, and how additional linking evidence can enable previously unavailable inferences. The authors evaluate their theoretical predictions through experiments on structured synthetic relational worlds using networks of three different depths (concepts for compositional).
Critical strengths
- The paper investigates an important question in compositional learning: how neural models learn to combine known relations into previously unseen inferences.
- The paper combines theoretical analysis with controlled experiments to investigate how compositional reasoning emerges during learning.
- The findings distinguish between compositional failure caused by slow learning and failure caused by insufficient evidence, even when the model has learned the premises. They also explain how additional linking evidence can enable previously unavailable inferences.
- The theory makes predictions about the learning trajectories and emergence times of individual compositional laws, going beyond aggregate performance.
- The authors evaluate these predictions across independently generated relational worlds and three network depths, providing evidence for the theory within the studied setting.
Critical weaknesses
- Although the findings are interesting, the analysis and experiments are limited to shallow and deep linear networks. The connection to nonlinear architectures, including Transformers used for autoregressive generation, remains untested.
- The experiments evaluate network depths of only one, two, and three layers. Although the theoretical formulation accommodates general depth, further evaluation would help establish whether its predictive accuracy persists in deeper networks.
- The evaluation focuses on learning continuous relational embeddings and retrieving a target entity from a predefined candidate set. Generalization to more complex prediction settings requires further investigation.
- All synthetic worlds use predefined additive relations. This formulation excludes noncommutative composition, where applying relation A followed by B differs from applying B followed by A, limiting the range of compositional structures covered by the study.
- For deeper networks, the authors integrate the derived dynamics from the same initial weights as the trained networks. The close agreement supports their approximation of training dynamics, but does not show that a simple spectral summary predicts emergence across architectures or settings.
Questions for authors
Please address the weaknesses raised above.
2 comments
Thank you to the authors for their responses. These responses resolved my initial confusion and concerns about the paper. The additional explanation better supports the theory and provides more comprehensive coverage. Based on my knowledge, they are convincing, so I am increasing my initial score.
We thank the reviewer for the constructive assessment and address each point below.
1. Nonlinear architectures. "The connection to nonlinear architectures, including Transformers... remains untested."
While our experiments use linear networks, our theory separates what depends on linearity from what holds across architectures (Section 6, Paragraph 4). Because Proposition 4.1 depends only on the solution set of AE=C, it determines identifiability and the effect of linking facts for any nonlinear model that can represent those embeddings (Section 4.2). By contrast, our timing predictions rely on linear dynamics, which have previously anticipated key learning phenomena in nonlinear networks and Transformers (Section 1, Paragraph 3). To test whether this spectral ordering carries over to nonlinear models directly, we outline an empirical test checking whether multi-hop compositions on knowledge-graph benchmarks (FB15k-237 and WN18RR) emerge in the order that the spectrum of their training facts predicts (Section 6, Paragraph 4).
2. Greater network depths (N=4 and N=5). "The experiments evaluate network depths of only one, two, and three layers."
We extended Experiment 1 to N=4 and N=5 on 30 worlds each, using the main experiment’s settings and no fitted parameters (Section 5.1, Paragraph 5; Appendix C.12, Table C.13). All 480 law-worlds per depth emerged in both the networks and the predictions, with median relative timing errors of 0.0% (N=4) and 0.5% (N=5) and trajectory agreement of R2=0.98 to 1.00. The theory also predicted the continued narrowing of the spread across laws, to 1.15-fold at N=4 (1.16 predicted) and 1.05-fold at N=5 (1.06 predicted). Agreement on individual emergence times is looser at these depths (R2 = 0.87 and 0.90), in part because the 16 laws now emerge within 5–15% of each other, which leaves little variance to explain.
3. Fixed candidate sets. "The evaluation focuses on learning continuous relational embeddings and retrieving a target entity from a predefined candidate set."
We agree that richer prediction settings, such as generative prediction, remain untested, and Section 6 (Paragraph 4) lists evaluation within a fixed candidate set as a limitation. Within our framework, however, the theory predicts the composed point itself, and the candidate set only determines how that point is scored (Section 3.3). Alongside retrieval accuracy, we report geometric error, the distance between the composed point and its true target in embedding space (Figure C.3). The theory predicts these continuous trajectories with R2≥0.997 in all experiments (Table C.1), and across 40 randomly generated designs, candidate spacing is uncorrelated with timing error (∣r∣≤0.08; Appendix C.9).
4. Non-commutative composition. "All synthetic worlds use predefined additive relations. This formulation excludes noncommutative composition..."
Our framework extends beyond additive laws in two ways (Section 6, Paragraph 5; Appendix A.7). Commuting multiplicative relations reduce directly to our linear setting after taking logarithms. Non-commutative matrix relations (Rz=RyRx) make the constraints bilinear, so the consistent solutions form a curved manifold. Linearizing around a consistent solution θ∗ shows that the null space of the constraint Jacobian J(θ∗) governs local identifiability, while the spectrum of J(θ∗)⊤J(θ∗) sets local learning rates (Rothenberg, 1971). Developing a global dynamical theory for non-commutative relations remains an open problem.
5. Spectral summary vs. numerical integration. "The close agreement supports their approximation of training dynamics, but does not show that a simple spectral summary predicts emergence across architectures or settings."
To isolate what a spectral summary predicts without integrating coupled layers from initial weights, we decomposed the deep dynamics in the eigenbasis of H (Section 4.4; Appendix A.5). At N=1, emergence follows in closed form from the eigenvalues of H, the law’s loadings on them and the initial errors (Proposition 4.2). At N≥2, each isolated mode grows at a rate set by ck=λksk=∣∣qk⊤B∣∣2, which depends only on the training constraints A and C (Table C.6). This summary alone reproduces the compression of emergence times from 5.4-fold at N=1 to 1.8-fold at N=2, close to the networks’ 2.1-fold (Section 5.1, Paragraph 4; Table C.7). What it leaves out is the interaction between modes through the shared d×d layers. Without it, the summary overestimates emergence times 2.8-fold at N=2 and 7.1-fold at N=3, and misses the further compression to 1.2-fold at N=3 (Tables C.7, C.8). Individual emergence times at depth therefore require the full dynamics.