When Does Routing Become Interpretable? Causal Probes on Block Attention Residuals

• URL: http://arxiv.org/abs/2606.13168v1
• Date: Thu, 11 Jun 2026 10:37:04 GMT
• Title: When Does Routing Become Interpretable? Causal Probes on Block Attention Residuals
• Abstract summary: Block Attention Residuals (Block AttnRes) replace fixed additive residuals with a learned softmax over earlier depth-source representations.<n>We ask whether such exposure suffices for mechanistic interpretation.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Block Attention Residuals (Block AttnRes) by replace fixed additive residuals with a learned softmax over earlier depth-source representations, surfacing cross-layer routing as an inspectable tensor in the forward pass. This is a tempting interpretability target: information flow normally inferred indirectly is now directly observable. We ask whether such exposure suffices for mechanistic interpretation. We probe two same-scale ($0.6$B) Block AttnRes checkpoints under identical routing-ablation interventions: a vanilla Qwen3 inference-wrapped through a deterministic recency-bias schedule that the codebase admits as a routing-equivalent loading path, and a Block AttnRes Qwen3 trained from scratch with routing as part of optimisation. The wrapped baseline's routing weights are content-independent and reproduce the schedule's analytic prediction. The trained AttnRes checkpoint instead exhibits three localised routing motifs: an embedding-source pathway through early-layer MLP, a current-state pathway through early-layer attention and MLP, and an older-history pathway through late-layer attention. Beyond this stratification, we find a sharp dissociation between average routing mass and causal importance: in both sublayers, the largest mass slice is not the largest causal contribution, and one source family carries appreciable mass with no detectable causal role under intervention. Architectural exposure of routing is therefore necessary but not sufficient for mechanistic interpretation: structured depth routing emerges only when routing has been part of training, and even then, descriptive routing summaries should be treated as candidate hypotheses to be tested by causal interventions, not as evidence of mechanism in their own right.

Related papers

• Revisiting Complete Reasoning Traces for Post-Training [61.76889440384448]
  Large language models (LLMs) are often post-trained on pre-collected reasoning trajectories to improve their reasoning capability.<n>We show that full trajectories provide only limited benefit, while partial trajectories are effective even under heavy truncation.
  arXiv  Detail & Related papers  (2026-09-07T06:46:17Z)
• CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing [90.63350360476294]
  We present CRISP (Cliff-awaRe Input-adaptive Sparse Prefilling), which identifies and addresses two structural challenges in this dynamic routing paradigm.<n>We show that the routing decision can be read directly off the structure of the proxy attention map.<n>We formalize the post-softmax mass cliff and demonstrate theoretically that strictly cumulative coverage thresholds accumulate O(n) background noise at long contexts.
  arXiv  Detail & Related papers  (2026-09-01T22:49:46Z)
• Routing Divergence Is Not Evidence of Behavioral Influence in Same-Weight MoE Self-Distillation [7.901904409617788]
  Two Mixture-of-Experts (MoE) forward passes can share every weight yet route the same token through different experts.<n>We study this mismatch in its single-step form, with frozen weights rather than as a proxy for a full training trajectory.
  arXiv  Detail & Related papers  (2026-08-16T15:07:17Z)
• How Does Reasoning Flow? Tracing Attention-Induced Information Flow for Targeted RL in LLMs [52.712343092801326]
  FlowTracer is an RL framework that traces answer-targeted reasoning flow on an attention-weighted directed acyclic graph.<n>It scores tokens by flow throughput, revealing high-impact hubs and aggregation checkpoints that mediate long-range dependencies.<n>These derived importances are used to shape token-level rewards, enabling learning signals to focus precisely on the tokens that route information toward correct answers.
  arXiv  Detail & Related papers  (2026-06-09T09:56:51Z)
• From Sampled Outcomes to Capability Distributions: Rethinking Supervision for LLM Routing [79.19755531338872]
  Existing routing methods treat a model's single response to a query as its capability label for training.<n>We show that this assumption introduces systematic noise into routing supervision, making learned routing policies less reliable.<n>We propose DARS (Distribution-Aware Routing Supervision), a framework that constructs routing supervision from a distributional view of model behavior.
  arXiv  Detail & Related papers  (2026-06-05T05:42:00Z)
• Meta-Attention: Bayesian Per-Token Routing for Efficient Transformer Inference [0.0]
  We propose Meta-Attention, a framework that dynamically routes each token to the most appropriate attention strategy.<n>Unlike prior routing approaches that use deterministic or prior-free learned routing, the Meta-Controller treats per-token mechanism selection as posterior inference.<n>This design produces principled routing uncertainty estimates that govern the soft-to-hard routing transition, mitigates routing collapse without ad hoc load-balancing losses, and yields better compute-performance trade-offs.
  arXiv  Detail & Related papers  (2026-05-27T12:21:28Z)
• Route Before Retrieve: Activating Latent Routing Abilities of LLMs for RAG vs. Long-Context Selection [57.3886742625188]
  Pre-Route is a proactive routing framework that performs structured reasoning before answering.<n>Our study shows three key findings: (i) LLMs possess latent routing ability that can be reliably elicited with guidelines; (ii) linear probes reveal that structured prompts sharpen the separability of the "optimal routing" in representation space; and (iii) distillation transfers this reasoning structure to smaller models for lightweight deployment.
  arXiv  Detail & Related papers  (2026-05-11T09:10:55Z)
• How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models [0.0]
  This paper localizes the policy routing mechanism in alignment-trained language models.<n>An intermediate-layer attention gate reads detected content and triggers deeper heads that boost the signal toward refusal.
  arXiv  Detail & Related papers  (2026-04-06T03:20:37Z)
• Route-Induced Density and Stability (RIDE): Controlled Intervention and Mechanism Analysis of Routing-Style Meta Prompts on LLM Internal States [10.639846833695806]
  We quantify internal density via activation sparsity, (C2) domain-keyword attention, and (C3) output stability via predictive entropy and semantic variation.<n>We present RIDE as a diagnostic probe for calibrating routing design and uncertainty estimation.
  arXiv  Detail & Related papers  (2026-03-31T03:19:36Z)
• Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures [58.54426802984356]
  We introduce a causal evaluation protocol that makes explicit intermediate structures causally determine the output.<n>Across eight models and three benchmarks, models appear self-consistent with their own intermediate structures but fail to update predictions after intervention in up to 60% of cases.<n>Overall, intermediate structures in schema-guided pipelines function as influential context rather than stable causal mediators.
  arXiv  Detail & Related papers  (2026-03-17T13:01:44Z)
• TraceRouter: Robust Safety for Large Foundation Models via Path-Level Intervention [44.64827167753535]
  harmful semantics act as distributed, cross-layer circuits, rendering localized interventions brittle and detrimental to utility.<n>We propose textbfTrace, a path-level framework that traces and disconnects the causal propagation circuits of illicit semantics.<n>Trace significantly outperforms state-of-the-art baselines, achieving a superior trade-off between adversarial robustness and general utility.
  arXiv  Detail & Related papers  (2026-01-29T15:58:12Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.