What happened
arXiv:2609.17804v1 Announce Type: new Abstract: Large language models solve grade-school math word problems with high accuracy, yet a single irrelevant clause inserted into the problem can collapse it. We reconcile these observations with a mechanistic account. We show that the model's internal computation decomposes into a four-stage sequential pipeline, Schema Abstraction, Operation Planning, Operand Binding, and Computation, each stage producing a distinct intermediate representation in an identifiable band of layers. Using the same scaffold to diagnose distractor-induced failure, we localize the corruption to a single stage, Operation Planning, implemented by a set of attention heads whose causal role we validate bidirectionally. In short, we provide a mechanistic interpretation of math word problem reasoning in LLMs, and their failure when distracted. Update (2026-09-15): Show HN: Capsule – Single-file web apps that save their data into SQLite. Hey HN, I always had the problem that building HTML pages is really simple now, but trying to save data required hosting it somewhere, and sharing it afterwards was not easy. Over the last few months, I've been building an app called Capsule (it’s also the file extension name) written in Rust with Tauri 2.0 that allows packing an HTML app and its data into a single SQLite file. The HTML file and any related assets are directly embedded in the database. User data can either be saved as a localStorage key/value store or via a MongoDB-inspired collections API as documents, saved in a table in the file. You can also save other assets, like PDF files or images, directly in the database to keep different documents together. All data can be easily exported to CSV or JSON if needed. Privacy and security were a big priority for me, so documents cannot do anything out of the box. They don’t have direct access to the file system and they require permission to access the internet. The permission model is still something I’m working to improve. Capsule documents can also use local or remote AI model Update (2026-09-21): TinyCeNN-LM: Quality-Gated Conversion of Pretrained Attention with CeNN-Inspired Cellular-Recurrent Layers. arXiv:2609.21139v1 Announce Type: new Abstract: Replacing attention in a pretrained language model is a compatibility problem: a plausible substitute may alter representations expected by later layers. TinyCeNN-LM introduces a \emph{quality-gated post-training conversion} framework using CeNN-inspired cellular-recurrent layers with bounded local processing, compact recurrent memory, routing, fusion, and accept-or-rollback validation. Three implementations are studied: Integrated Memory, MemoryFusion, and PDelta3-GDN2-CLVR+Local32. Strict PDelta3 conversion accepts a layer only when representation and NLL criteria pass fixed thresholds. On SmolLM2-135M, layers 0-2 are accepted with cumulative $\Delta\mathrm{NLL}=+0.01209$, while layer 3 is rejected despite acceptable NLL because representation fidelity fails. On Qwen3.5-0.8B, full-attention layers 3, 7, and 11 are accepted with final $\Delta\mathrm{NLL}=+0.02073$. Integrated Memory keeps perplexity within $-0.07\%$ to $+0.93\%$ while reducing total cache by up to $6.01\%$. A sampled 200-item downstream Update (2026-09-23): Spectra: A Rules-Driven LLM Pipeline for Automated KYC Document Processing. arXiv:2609.25474v1 Announce Type: new Abstract: Know Your Client (KYC) onboarding in capital markets requires analysts to manually classify documents, extract structured data from heterogeneous sources, and validate compliance against complex regulatory policies. This process requires significant analyst time per client, with end-to-end onboarding often stretching to multiple weeks due to sequential handoffs. In this work, we analyze an on-boarding process and find that it comprises repeatable components well-suited to AI automation. We therefore propose a restructured workflow to be amenable to automation: we consolidate the traditional four-party process into two parties that share most of the work and can be automated together, eliminating intermediate handoffs that compound delays. To automate the remaining steps, we introduce Spectra, an AI-assisted document processing platform that combines a structured rules engine with LLM-based classification, extraction, and validation agents. The rules engine encodes compliance policy as a queryable database, enabling focused context injec