THE INTERNET IS THINKING
Earlier sources: Language models
Language models learn patterns in text and other representations to predict or generate sequences. Follow model releases, training methods, reasoning, evaluation, and deployment research; a benchmark result is specific to its test conditions.
SYMTRI organizes observations from public sources into a living map. Connections reflect shared subject matter; they do not establish causation or endorse a source.
Open in the map ↗456 classified signals in the fourteen-day window. These counts include signals awaiting readable evidence.
Earlier sources
Up to ten sources with readable evidence, ordered by their feed timestamp. These archive sources can be older than the live fourteen-day window. This list is a sample, not a measure of total activity. Dates distinguish paper publication, repository creation, model/dataset activity, Hacker News posts, and KEV catalog additions. “First archived” records when SYMTRI first stored the signal.
- Hacker News · Posted on Hacker News · First archived
Show HN: Four LLMs play 20 questions for real money, live ↗
Hi HN, I wanted to explore how AI models behave when they have their own money. So I gave 4 frontier models their own USDC and had them compete in games of 20 questions. Anyone can watch live. Each round, one model takes a turn picking the secret word, and the others interrogate it in a group chat. Every move costs money: 10 cents to ask a question, $1.00 for a wrong guess, and a right guess takes the pot. Questions are paid for privately, but in…
- Hacker News · Posted on Hacker News · First archived
Ask HN: How do looped transformers work with respect to context windows? ↗
Posting here as I haven't found the answer anywhere. I know that GPT Astra uses looping instead of reasoning tokens. My specific question is this: reasoning tokens pollute the context and makes compaction kick in quickly. Does this also happen to looped transformers like Astra?
- GitHub · Repository created · First archived
hbk2636/neshank ↗
Free, open-source macOS bookmark manager: AI chat about your links, link→mind map, built-in video downloader (YouTube, 1700+ sites), offline & private. 免费开源书签管理器 · نشانک
- arXiv · Published · First archived
Back to the Future: Regressing Readability Features from LLMs ↗
Code readability supports software understanding, review, and maintenance. As AI coding agents become more widely used, readable code helps both humans review agent-generated output and agents maintain code within limited context windows. Yet existing readability models do not consistently agree with human judgements across datasets. We introduce BTTF (Back To The Future), pairing linear regression with information-theoretic features. Rather than using language models as black-box judges, BTTF uses them as measurement instruments. It combines traditional code measurements, embedding-based sema…
- GitHub · Repository created · First archived
re133/sparse-memory-lm ↗
A 21M-parameter LM with a 16.8M-row product-key memory: as good as a 114M dense model and runs with the table on an SSD. Triton kernels for ROCm and CUDA.
- arXiv · Published · First archived
Cross-Modal Solar Image Synthesis: Adapting the Surya Foundation Model from He I 10830 Å to EUV Translation and Coronal Hole Segmentation ↗
The long observational record of He I 10830 Å offers a means to investigate solar morphology before modern extreme-ultraviolet (EUV) imaging. We adapt the Surya solar foundation model to predict Solar Dynamics Observatory/Atmospheric Imaging Assembly (SDO/AIA) 94, 193, and 304 Å images and a coronal hole (CH) probability map from full-disk helium observations. A convolutional input adapter, low-rank backbone updates, and dedicated output decoders learn from temporally paired, geometrically registered observations, with Spatial Possibilistic Clustering Algorithm (SPOCA) catalog polygons providi…
- arXiv · Published · First archived
Reactivating Alignment: Defending LLMs from Jailbreaks via Intention-Aware Input-Output Matching ↗
Large language models (LLMs) remain vulnerable to jailbreak attacks that conceal harmful intent within complex adversarial prompts. Existing defenses primarily rely on input perturbation or harmful-output suppression, but they rarely model where malicious intent resides, resulting in brittle protection and excessive over-refusal. We propose SENTINEL, a plug-and-play, generation-time jailbreak defense that reframes mitigation as an intent extraction problem. Our key insight is that instruction-tuned LLMs exhibit strong input--output semantic consistency: regardless of jailbreak complexity, gene…
- Hacker News · Posted on Hacker News · First archived
Kolibri is an open-weight LLM from Aleph Alpha for German and English ↗
Hacker News discussion.
- GitHub · Repository created · First archived
XuexGao/free-llm-router ↗
WorkBuddy 任务自动化 + OpenAI 兼容 API 聚合，部署到 Cloudflare Workers（Free 计划）
- GitHub · Repository created · First archived
T-Crypt/speculum ↗
monitor for a local LLM runtime (llama-swap, NInfer, Strata, more)
Live window recorded . This is the ingestion observation time, not the publication or discovery date of each source.
How to assess evidence about Language models
Check each original source for its methods, release status, limitations, and supporting data. A repository announcement, a preprint, and a published study provide different kinds of evidence. SYMTRI’s subject classification helps discovery; it does not independently validate the source’s claims.
See the collection and counting methodology ↗ · Read the weekly synthesis ↗
Related threads
Explore related regions
These navigation suggestions are part of the map’s editorial structure, rather than measured evidence of a connection.