The Archive

Dataset and design guidelines for assessing cultural alignment in LLMs, addressing limitations of prior cultural bias evaluation approaches.

Quantum annealing method for interpretable feature selection in CNNs applied to image classification.

Give Toothcomb a speech transcript and it will fact-check and analyse it. If you have an MP3 file of someone speaking, it can generate the transcript for you. You can also stream audio in real time from your device's microphone. You can see a [demo running here](https://toothcomb.codebox.net/) and read more about the project on the [home page](https://codebox.net/pages/toothcomb-ai-fact-checker). Analysis is performed in three stages: 1. The text is broken up into small parts, each usually a few sentences in length. These parts are sent, one at a time, to the Claude Opus API with [detailed...

Prefill-time intervention technique to reduce hallucinations in large vision-language models by addressing accumulation errors during decoding.

Experimental study demonstrating LLMs can be manipulated to prioritize fringe scientific material and generate misleading fluent responses contradicting scientific consensus.

Mandelbrot rank-frequency distribution identified across frontier LLM outputs enables sub-microsecond token verification, 100,000× faster than sampling-based detection.

Reddit user reports renewed enthusiasm for personal coding project after using Claude for 6 weeks.

Matthew Yglesias argues for AI-assisted professional software development over autonomous "vibe coding," prioritizing human-managed productivity gains.

HotComment: multimodal benchmark for evaluating online comment popularity across platforms using video, text, and content quality metrics.

Microsoft study identifies job categories most exposed to AI automation; labor market impact analysis.

Nonverbal Syntax Framework systematizes 908 studies mapping nonverbal behavioral cues to learner cognitive/affective states for adaptive education systems.

The startup specializes in "non-invasive" "mind-reading" tech—a kind of neural data collection that, its CEO hopes, will have all sorts of consumer applications.

Benchmark comparing abliteration techniques across GLM-4.7-Flash (MoE architecture) vs. prior Qwen family tests; evaluates HauhauCS uncensored claims.

WhisperPipe: streaming architecture for real-time ASR maintaining transcription accuracy with bounded memory through hybrid VAD and context management.

Semantic search system deployed at children's hospital indexing 166M clinical notes using instruction-tuned embeddings; addresses scalability and governance challenges.

OxyGent open-source framework enables modular, observable multi-agent systems via pluggable components and permission-driven dynamic planning.

Study examines LLM integration in hybrid work environments to adjust spatial experiences and collaboration dynamics.

Empirical study comparing PLM-GNN hybrids for code classification and vulnerability detection; hybrids outperform GNN-only baselines.

Reddit discussion speculating on potential end of discounted Claude subscription pricing models.

Tank OS puts OpenClaw AI agents into a container that let's it run reliably and more safely, especially for those running fleets of them.

Qwen3.6-27B IQ4_XS quantization bloat analysis; reverting llama.cpp commit reduces VRAM from 15.1GB to 14.7GB with 110k context.

Generic discussion post about wisdom or best practices in AI/coding communities.

First systematic study of uncertainty estimation in audio-aware LLMs; benchmarks five methods addressing hallucination and confidence calibration.

I was overcharged by more than $100, so I opened a billing ticket last month. They only responded yesterday and said everything looked fine because they refunded me $100 in credits. They didn’t give me any option to choose between a refund to my card or credits, but I can let that go... The worst part is what happened next: due to what seems like an error on their side, I lost access to my plan. I no longer have 5x Max and my account now shows as Free. This is insane. Do I really have to wait another month to fix this while not having access to the service I already paid for? My billing c...

DualFact multimodal framework separates factual verification in procedural video captioning into conceptual and contextual facts.

Analysis of Perspective API shutdown exposes structural dependence of NLP/LLM evaluation on single proprietary toxicity measurement tool.

Marco-MoE open-weight multilingual sparse MoE models with 5% parameter activation and best-in-class performance-to-compute ratio.

Dictionary learning method for Kernel EDMD approximation of nonlinear dynamical systems via Koopman operators.

RealMat-BaG benchmark for semiconductor bandgap prediction under experimental conditions using GNNs; addresses domain generalization challenges.

SnapGuard detects prompt injection attacks on screenshot-based web agents using lightweight multimodal methods instead of large VLMs.