The daily ledger

Don't classify. Hallucinate!

Don't classify. Hallucinate! I still have quite a bit of older content on my blog that I never got round to tagging. My blog has 1,856 tags - likely too many to feed to an LLM in one go and say "which of these tags…

How Claude’s text watermark works

In this article, we share answers to some of the questions we’ve received about how our chosen watermarking method works, whether it affects Claude’s outputs, and why we’re making this change.

The Codebase Is the Agent's Working Environment

Agents generate code in seconds, so investing in code quality looks obsolete. My daily work with agents shows the opposite. Every session starts by reading the existing code, and its quality decides how productive the…

State of Open Models: Summer 2026 Observations

Google's $99 Fitbit gets a wild new update

PLUS: Flock tightens data rules after backlash

OpenAI feels the frontier need for speed

PLUS: Build a work 'Second Brain' that updates itself

ViTOED: A Dataset for Target-Oriented Emotion Detection on Vietnamese Social Media Texts

arXiv:2608.12776v1 Announce Type: new Abstract: This paper introduces ViTOED, a novel dataset for target-oriented emotion detection in Vietnamese social media texts. The ViTOED comprises 10,985 user comments and 21,244…

Better Decomposition, Free Aggregation: A Synthesizer-Folding Framework for Multilingual Multi-Hop Question Answering

arXiv:2608.13160v1 Announce Type: new Abstract: Multilingual retrieval-augmented generation (mRAG) equips large language models with access to globally distributed external knowledge for complex multilingual question…

SAEVerbalizer: Generating Explanations for Sparse Autoencoder Features via Representation Verbalization

arXiv:2608.13538v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily…

HC-RAG: Evidence-Centric Retrieval-Augmented Generation over Heterogeneous Financial Filings

arXiv:2608.12335v1 Announce Type: new Abstract: Financial question answering over annual reports requires more than retrieving semantically similar passages. It often involves identifying relevant companies and fiscal…

CASA: Content-Acoustic Speaking Assessment with Speech Encoder and Large Language Model

arXiv:2608.13101v1 Announce Type: new Abstract: Research on automatic speaking assessment (ASA) has increasingly adopted multimodal speech large language models to assess learners' speaking performance. However,…

CityRiSE: Reasoning Urban Socio-Economic Status in Large Vision-Language Models via Reinforcement Learning

arXiv:2510.22282v2 Announce Type: replace-cross Abstract: Urban socio-economic sensing plays a vital role in advancing global sustainable development goals. With the advent of Large Vision-Language Models (LVLMs), new…

Natural Language Processing: A Comprehensive Practical Guide from Tokenisation to RLHF

arXiv:2605.03799v3 Announce Type: replace Abstract: This preprint presents a systematic, research-oriented practicum that guides the reader through the entire modern NLP pipeline --- from tokenisation and vectorisation…

Tracing Provenance and Detecting Tampering with Complementary LLM Watermarks

arXiv:2608.12713v1 Announce Type: cross Abstract: Watermarking LLM-generated text is an important task for tracing its provenance. Existing LLM watermarks preserve provenance under editing, but this same robustness…

Large Language Models Can Follow Instructions, But Not Many at Once: Phase Transitions in Compositional Constraint Satisfaction

arXiv:2608.12426v1 Announce Type: cross Abstract: Large language models are increasingly deployed in settings that require simultaneous adherence to multiple explicit constraints - reasoning structure, safety…

Self-Referential Induction Increases Response Instability Relative to Unresolvable and Verifiable Questions in Large Language Models

arXiv:2608.13258v1 Announce Type: new Abstract: Self-referential prompting has been shown to reliably induce large language models to produce first-person reports resembling subjective experience, but no prior work…

LigBench: A Unified and Human-Aligned Benchmark for LLM-based Research Idea Generation

arXiv:2608.13136v1 Announce Type: new Abstract: With the rapid advancement of large language models (LLMs), research idea generation has attracted increasing attention. Existing approaches enable LLMs to retrieve…

When Your Agent Opens the Chat App: Agent-Controlled Search over Raw Chat Logs Rivals Structured Memory

arXiv:2608.12888v1 Announce Type: new Abstract: Agent-memory systems increasingly buy retrieval quality with structure, transforming raw conversation histories into summaries, embeddings, trees, or knowledge graphs…

Evaluation design conditions the expert-vs-auto MeSH gap: a controlled comparison of bag-of-words and BiomedBERT on the Cohen benchmark

arXiv:2607.21685v3 Announce Type: replace Abstract: A systematic review begins with someone reading thousands of abstracts to identify the few that are relevant, and classifiers are used to prioritise that reading.…

The Evaluator Is Part of the Experiment: Measuring Open-Ended LLM Conformity

arXiv:2608.04463v2 Announce Type: replace Abstract: Prior work on LLM conformity largely measures discrete answer flips under verifiable labels. Open-ended revisions require a different measurement strategy because…