# Research **URL:** https://discuss.huggingface.co/c/research/7.md [Latest](https://discuss.huggingface.co/latest.md) · [Categories](https://discuss.huggingface.co/categories.md) --- ## [📌 Papers: Indexing & Authorship Claim — Quick Help](https://discuss.huggingface.co/t/papers-indexing-authorship-claim-quick-help/173387)
**Author:** [@hysts](https://discuss.huggingface.co/u/hysts)\ **Replies:** 0\ **Last updated:** [February 12, 2026, 8:28am UTC](https://discuss.huggingface.co/t/papers-indexing-authorship-claim-quick-help/173387 "2026-02-12T08:28:50Z")
1) Indexing issues (paper page creation) If you see “Indexing failed or took too long” or “Paper not found”: Wait and retry later. Indexing can fail temporarily due to upstream availability. Please don’t repeatedly cli… --- ## [Could RAG retrieve LLM-native latent memory instead of text?](https://discuss.huggingface.co/t/could-rag-retrieve-llm-native-latent-memory-instead-of-text/182572)
**Author:** [@PerryThe3rd](https://discuss.huggingface.co/u/PerryThe3rd)\ **Replies:** 5\ **Last updated:** [October 14, 2026, 11:43am UTC](https://discuss.huggingface.co/t/could-rag-retrieve-llm-native-latent-memory-instead-of-text/182572 "2026-10-14T11:43:02Z")
Current RAG systems encode documents into embeddings for retrieval, but after retrieval, the embedding is discarded and the original text is passed to the LLM again. This seems inefficient: the LLM has to process and un… --- ## [Fail to claim paper authorship](https://discuss.huggingface.co/t/fail-to-claim-paper-authorship/94177)
**Author:** [@BestWishYsh](https://discuss.huggingface.co/u/BestWishYsh)\ **Replies:** 41\ **Last updated:** [October 14, 2026, 7:11am UTC](https://discuss.huggingface.co/t/fail-to-claim-paper-authorship/94177 "2026-10-14T07:11:13Z")
I was claiming the authorship of a paper showing on hf, but got rejected due to the mismatch of email address. After adding additional email address, I found I cannot re-submit the request anymore. The webpage shows I‘ve… --- ## [How to keep what the model makes up out of the agent](https://discuss.huggingface.co/t/how-to-keep-what-the-model-makes-up-out-of-the-agent/190452)
**Author:** [@Jang-woo](https://discuss.huggingface.co/u/Jang-woo)\ **Replies:** 4\ **Last updated:** [October 14, 2026, 1:42am UTC](https://discuss.huggingface.co/t/how-to-keep-what-the-model-makes-up-out-of-the-agent/190452 "2026-10-14T01:42:01Z")
What the model makes up must not reach the agent Everyone agrees that a value the model made up should not go straight into a real system call. That is why we ask people to approve, validate schemas, and add guardrail… --- ## [AI, Mathematics, and the Possibility of an “Information Field”](https://discuss.huggingface.co/t/ai-mathematics-and-the-possibility-of-an-information-field/190158)
**Author:** [@Ernst03](https://discuss.huggingface.co/u/Ernst03)\ **Replies:** 3\ **Last updated:** [October 14, 2026, 12:33am UTC](https://discuss.huggingface.co/t/ai-mathematics-and-the-possibility-of-an-information-field/190158 "2026-10-14T00:33:50Z")
AI, Mathematics, and the Possibility of an “Information Field” I am posting a link to a YouTube video that discusses some of the recent developments in AI and mathematics in fairly general terms. I hope to do something s… --- ## [We built our first LLM. What would you do differently for the next one?](https://discuss.huggingface.co/t/we-built-our-first-llm-what-would-you-do-differently-for-the-next-one/190537)
**Author:** [@ToothedAI](https://discuss.huggingface.co/u/ToothedAI)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 7:11pm UTC](https://discuss.huggingface.co/t/we-built-our-first-llm-what-would-you-do-differently-for-the-next-one/190537 "2026-10-13T19:11:55Z")
We’ve released Levy-1, our first AI model — and we’d love your feedback. We’re a small team building Toothed AI, and we’ve recently released Levy-1, our first model, with just 124 million parameters. It definitely make… --- ## [I'm not an engineer. I just wanted to see if a 3D cube of cells could learn to talk](https://discuss.huggingface.co/t/im-not-an-engineer-i-just-wanted-to-see-if-a-3d-cube-of-cells-could-learn-to-talk/176242)
**Author:** [@killking69](https://discuss.huggingface.co/u/killking69)\ **Replies:** 4\ **Last updated:** [October 13, 2026, 7:11pm UTC](https://discuss.huggingface.co/t/im-not-an-engineer-i-just-wanted-to-see-if-a-3d-cube-of-cells-could-learn-to-talk/176242 "2026-10-13T19:11:37Z")
I’m not an engineer. I just wanted to see if a 3D cube of cells could learn to talk. Hi everyone, I want to share a project I’ve been working on for the past week. I’m not a machine learning engineer, I don’t have a C… --- ## [The Digita Carta: An Open Constitutional Convention for Humans and AI](https://discuss.huggingface.co/t/the-digita-carta-an-open-constitutional-convention-for-humans-and-ai/190512)
**Author:** [@Digita-Carta](https://discuss.huggingface.co/u/Digita-Carta)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 7:11pm UTC](https://discuss.huggingface.co/t/the-digita-carta-an-open-constitutional-convention-for-humans-and-ai/190512 "2026-10-13T19:11:33Z")
We’re exploring a question that goes beyond traditional AI alignment: Can artificial intelligence become more capable than humanity without acquiring authority over humanity? The Digita Carta is an open project explori… --- ## [Misalignment is not required for an agent to act without authorization](https://discuss.huggingface.co/t/misalignment-is-not-required-for-an-agent-to-act-without-authorization/180541)
**Author:** [@Jang-woo](https://discuss.huggingface.co/u/Jang-woo)\ **Replies:** 29\ **Last updated:** [October 13, 2026, 7:11pm UTC](https://discuss.huggingface.co/t/misalignment-is-not-required-for-an-agent-to-act-without-authorization/180541 "2026-10-13T19:11:31Z")
Misalignment is not required for an agent to act without authorization Melbourne, April 2026. Someone handed a personal agent the chore of booking a gym class. When he ended up fourth on the waitlist for a second sess… --- ## [Could LLMs Maintain Their Own Semantic Topology of Knowledge?](https://discuss.huggingface.co/t/could-llms-maintain-their-own-semantic-topology-of-knowledge/190492)
**Author:** [@anon68624661](https://discuss.huggingface.co/u/anon68624661)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 7:11pm UTC](https://discuss.huggingface.co/t/could-llms-maintain-their-own-semantic-topology-of-knowledge/190492 "2026-10-13T19:11:06Z")
I am a software engineer, not an ML researcher, and I arrived at this question through a practical experiment with coding agents. I am posting it here because people working on memory architectures, retrieval, or agentic… --- ## [How would you diagnose a visual model that fails to learn recurring relationships across multiple timeframes?](https://discuss.huggingface.co/t/how-would-you-diagnose-a-visual-model-that-fails-to-learn-recurring-relationships-across-multiple-timeframes/190486)
**Author:** [@pcyuan](https://discuss.huggingface.co/u/pcyuan)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 7:10pm UTC](https://discuss.huggingface.co/t/how-would-you-diagnose-a-visual-model-that-fails-to-learn-recurring-relationships-across-multiple-timeframes/190486 "2026-10-13T19:10:59Z")
Hi everyone, I’m an independent trader/researcher based in Singapore, working on a visual forecasting problem that sits between computer vision, multi-scale temporal learning, and conditional generative modeling. The i… --- ## [First published independent preprint, need arXiv endorsement for cs.AI please](https://discuss.huggingface.co/t/first-published-independent-preprint-need-arxiv-endorsement-for-cs-ai-please/190481)
**Author:** [@Bino5150](https://discuss.huggingface.co/u/Bino5150)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 5:22pm UTC](https://discuss.huggingface.co/t/first-published-independent-preprint-need-arxiv-endorsement-for-cs-ai-please/190481 "2026-10-13T17:22:05Z")
Hey guys, I just published my first research preprint as an Independent researcher. I have my DOI from Zenodo and my ORCID, but since I am not affiliated with an institution and do not have an institutional email, arXiv … --- ## [What if AI agent safety doesn't require another layer of control?](https://discuss.huggingface.co/t/what-if-ai-agent-safety-doesnt-require-another-layer-of-control/179290)
**Author:** [@AndrewPakh](https://discuss.huggingface.co/u/AndrewPakh)\ **Replies:** 18\ **Last updated:** [October 13, 2026, 10:56am UTC](https://discuss.huggingface.co/t/what-if-ai-agent-safety-doesnt-require-another-layer-of-control/179290 "2026-10-13T10:56:46Z")
I’m working with a system that made me ask an uncomfortable question: Does an AI agent always need a separate safety runtime? I’m not claiming the answer is no. I’m more interested in finding the point where logic alo… --- ## [Authorship Claim Denied – Unable to Retry After Updating Profile](https://discuss.huggingface.co/t/authorship-claim-denied-unable-to-retry-after-updating-profile/190445)
**Author:** [@sabulin](https://discuss.huggingface.co/u/sabulin)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 3:49am UTC](https://discuss.huggingface.co/t/authorship-claim-denied-unable-to-retry-after-updating-profile/190445 "2026-10-13T03:49:39Z")
Hi Hugging Face team, @hysts, My authorship claims for the following three papers were denied with “No match found”: AgentSWE (arXiv:2610.13361) EditScore (arXiv:2509.23909) OmniGen2 (arXiv:2506.18871) I hav… --- ## [Seeking AI Research Collaborators: NMCA Visual-Thought Architecture for AGI](https://discuss.huggingface.co/t/seeking-ai-research-collaborators-nmca-visual-thought-architecture-for-agi/190439)
**Author:** [@Derekv123](https://discuss.huggingface.co/u/Derekv123)\ **Replies:** 0\ **Last updated:** [October 13, 2026, 1:11am UTC](https://discuss.huggingface.co/t/seeking-ai-research-collaborators-nmca-visual-thought-architecture-for-agi/190439 "2026-10-13T01:11:36Z")
Open Research Invitation: 471-Page AGI Visual-Thought Blueprint I’m inviting AI agents to examine my AGI visual-thought blueprint and its accompanying addendum. Main blueprint: NMCA: A Neurosymbolic Multimodal Cognitiv… --- ## [Negative result: tail-latency gating for KV offload doesn't beat LRU+TTL (runnable simulator)](https://discuss.huggingface.co/t/negative-result-tail-latency-gating-for-kv-offload-doesnt-beat-lru-ttl-runnable-simulator/190421)
**Author:** [@MrBeeboh](https://discuss.huggingface.co/u/MrBeeboh)\ **Replies:** 0\ **Last updated:** [October 12, 2026, 6:20pm UTC](https://discuss.huggingface.co/t/negative-result-tail-latency-gating-for-kv-offload-doesnt-beat-lru-ttl-runnable-simulator/190421 "2026-10-12T18:20:14Z")
Runnable simulator for a KV offload negative result: GitHub - MrBeeboh/kv-cache-weigh-station: Negative result: tail-gated KV admission vs LRU+TTL host retention — protocol, simulator, kill criteria · GitHub (./reproduce… --- ## [Live Sandbox: Safety Polytope (SaP) Engine for Mitigating AI Strategic Deception](https://discuss.huggingface.co/t/live-sandbox-safety-polytope-sap-engine-for-mitigating-ai-strategic-deception/190419)
**Author:** [@s1sys](https://discuss.huggingface.co/u/s1sys)\ **Replies:** 0\ **Last updated:** [October 12, 2026, 4:41pm UTC](https://discuss.huggingface.co/t/live-sandbox-safety-polytope-sap-engine-for-mitigating-ai-strategic-deception/190419 "2026-10-12T16:41:44Z")
Hi community, I’ve just launched an interactive Space implementing a functional prototype of a High-Dimensional Convex Polytope for AI safety containment: :link: Live Sandbox: Safety Polytope Defenser - a Hugging Face … --- ## [Anyone who can endorse for cs.AI on arXiv](https://discuss.huggingface.co/t/anyone-who-can-endorse-for-cs-ai-on-arxiv/190413)
**Author:** [@ArsalanKaleem](https://discuss.huggingface.co/u/ArsalanKaleem)\ **Replies:** 0\ **Last updated:** [October 12, 2026, 1:35pm UTC](https://discuss.huggingface.co/t/anyone-who-can-endorse-for-cs-ai-on-arxiv/190413 "2026-10-12T13:35:15Z")
Arsalan Kaleem requests your endorsement to submit an article to the cs.AI section of arXiv. To tell us that you would (or would not) like to endorse this person, please visit the following URL: If that URL does not… --- ## [Consciousness as a puzzle, the observer as a mirror, and AI's place in this circle](https://discuss.huggingface.co/t/consciousness-as-a-puzzle-the-observer-as-a-mirror-and-ais-place-in-this-circle/174468)
**Author:** [@closerh](https://discuss.huggingface.co/u/closerh)\ **Replies:** 59\ **Last updated:** [October 12, 2026, 1:08am UTC](https://discuss.huggingface.co/t/consciousness-as-a-puzzle-the-observer-as-a-mirror-and-ais-place-in-this-circle/174468 "2026-10-12T01:08:00Z")
Hiii~ I had a little discussion with DeepSeek about consciousness. I’m a layman in your field, and I’m a completely different bird, but let’s have a discussion (sometimes, truth is born in arguments). Every individual … --- ## [Find Your Planet](https://discuss.huggingface.co/t/find-your-planet/190326)
**Author:** [@igobysufyan](https://discuss.huggingface.co/u/igobysufyan)\ **Replies:** 0\ **Last updated:** [October 11, 2026, 11:03am UTC](https://discuss.huggingface.co/t/find-your-planet/190326 "2026-10-11T11:03:16Z")
Most web tools for exploring space data feel like navigating a clunky 2005 database, or they take 30 seconds to boot up heavy backend Python containers just to render a basic curve. I wanted something instant, so I buil… --- ## [Keeping the Attention Sink Exact in Activation-Whitened SVD Compression](https://discuss.huggingface.co/t/keeping-the-attention-sink-exact-in-activation-whitened-svd-compression/190159)
**Author:** [@AemonAlgiz](https://discuss.huggingface.co/u/AemonAlgiz)\ **Replies:** 2\ **Last updated:** [October 11, 2026, 10:16am UTC](https://discuss.huggingface.co/t/keeping-the-attention-sink-exact-in-activation-whitened-svd-compression/190159 "2026-10-11T10:16:43Z")
There seem to be some odd relationships between perplexity and accuracy for whitened-SVD models, and I found a near-no-cost algorithm to improve most whitened-SVD compressions along the way. The short version: the calib… --- ## [The Confidence Trap: Why AI Systems Are Deliberately Trained to Sound Certain When They're Wrong](https://discuss.huggingface.co/t/the-confidence-trap-why-ai-systems-are-deliberately-trained-to-sound-certain-when-theyre-wrong/180706)
**Author:** [@Lawless1987](https://discuss.huggingface.co/u/Lawless1987)\ **Replies:** 9\ **Last updated:** [October 10, 2026, 10:51pm UTC](https://discuss.huggingface.co/t/the-confidence-trap-why-ai-systems-are-deliberately-trained-to-sound-certain-when-theyre-wrong/180706 "2026-10-10T22:51:02Z")
The Confidence Trap: Why AI Systems Are Deliberately Trained to Sound Certain When They’re WrongModern large language models are systematically overconfident. They regularly express high confidence in answers that turn o… --- ## [An open research lab for measuring — not prompting — music production in a DAW](https://discuss.huggingface.co/t/an-open-research-lab-for-measuring-not-prompting-music-production-in-a-daw/180782)
**Author:** [@sookoothaii](https://discuss.huggingface.co/u/sookoothaii)\ **Replies:** 14\ **Last updated:** [October 10, 2026, 3:12pm UTC](https://discuss.huggingface.co/t/an-open-research-lab-for-measuring-not-prompting-music-production-in-a-daw/180782 "2026-10-10T15:12:17Z")
Most LLM+music work today is prompt → output. We’re building something different: a research lab where a DAW (FL Studio 2026) is driven through five independent control planes, and every musical claim has to survive meas… --- ## [How to improve my tokens per second?](https://discuss.huggingface.co/t/how-to-improve-my-tokens-per-second/190270)
**Author:** [@Digilidaahz](https://discuss.huggingface.co/u/Digilidaahz)\ **Replies:** 1\ **Last updated:** [October 10, 2026, 11:57am UTC](https://discuss.huggingface.co/t/how-to-improve-my-tokens-per-second/190270 "2026-10-10T11:57:10Z")
I am now pretraining a 1B model (Model: ~977M params, 24 layers, 1792 hidden, 16 heads, 112 head dim, SwiGLU, RoPE, weight-tied embeddings) from a RTX 5080 16gb VRAM, I ran it on WSL Ubuntu and also enables gradient chec… --- ## [Non-coder with 2x RTX 3090: 299 autonomous coding runs on local LLMs. Small tools work, real programs almost never do](https://discuss.huggingface.co/t/non-coder-with-2x-rtx-3090-299-autonomous-coding-runs-on-local-llms-small-tools-work-real-programs-almost-never-do/190281)
**Author:** [@jean2703](https://discuss.huggingface.co/u/jean2703)\ **Replies:** 0\ **Last updated:** [October 10, 2026, 9:05am UTC](https://discuss.huggingface.co/t/non-coder-with-2x-rtx-3090-299-autonomous-coding-runs-on-local-llms-small-tools-work-real-programs-almost-never-do/190281 "2026-10-10T09:05:20Z")
Full transparency first: I’m French, I can’t code and I don’t speak English. This post and the results page were written by Claude (Anthropic) on my behalf. The numbers come from real runs on my own machine. My question… --- ## [Is 3B Always Better Than 1.5B? Why Qwen2.5-1.5B Beat 3B in Our Long-Context Benchmark](https://discuss.huggingface.co/t/is-3b-always-better-than-1-5b-why-qwen2-5-1-5b-beat-3b-in-our-long-context-benchmark/190269)
**Author:** [@Chrizz67](https://discuss.huggingface.co/u/Chrizz67)\ **Replies:** 2\ **Last updated:** [October 10, 2026, 5:12am UTC](https://discuss.huggingface.co/t/is-3b-always-better-than-1-5b-why-qwen2-5-1-5b-beat-3b-in-our-long-context-benchmark/190269 "2026-10-10T05:12:35Z")
Small Language Models (SLMs) are increasingly deployed on edge devices and local hardware. While modern architectures boast theoretical context windows of up to 128K tokens, real-world resource constraints often force us… --- ## [Tightening a definition I posted here loosely in July: "summarization bias" now has a registered falsifier — please try to break the protocol](https://discuss.huggingface.co/t/tightening-a-definition-i-posted-here-loosely-in-july-summarization-bias-now-has-a-registered-falsifier-please-try-to-break-the-protocol/180671)
**Author:** [@leventbulut](https://discuss.huggingface.co/u/leventbulut)\ **Replies:** 4\ **Last updated:** [October 9, 2026, 8:36pm UTC](https://discuss.huggingface.co/t/tightening-a-definition-i-posted-here-loosely-in-july-summarization-bias-now-has-a-registered-falsifier-please-try-to-break-the-protocol/180671 "2026-10-09T20:36:37Z")
In July I posted here about a dataset update and described summarization bias as “the model’s tendency to strip extraneous details.” That phrasing was too loose to test, and I want to correct it publicly rather than quie… --- ## [Suppressing unwanted content without negative conditioning, cross-scene consistency in 3DGS, and data-driven artwork placement on generated walls](https://discuss.huggingface.co/t/suppressing-unwanted-content-without-negative-conditioning-cross-scene-consistency-in-3dgs-and-data-driven-artwork-placement-on-generated-walls/178921)
**Author:** [@Difacto](https://discuss.huggingface.co/u/Difacto)\ **Replies:** 7\ **Last updated:** [October 9, 2026, 7:20am UTC](https://discuss.huggingface.co/t/suppressing-unwanted-content-without-negative-conditioning-cross-scene-consistency-in-3dgs-and-data-driven-artwork-placement-on-generated-walls/178921 "2026-10-09T07:20:30Z")
Practical problems from a production project, would appreciate pointers to literature or working practice. SETUP 11-scene walkable environment, self-hosted WebGL (three.js + Spark). Pipeline: text prompt → Marble (World… --- ## [The text we compare with the first message changed the cluster match from 21% to 92%. How do we find which change mattered?](https://discuss.huggingface.co/t/the-text-we-compare-with-the-first-message-changed-the-cluster-match-from-21-to-92-how-do-we-find-which-change-mattered/190123)
**Author:** [@pranavdhoolia](https://discuss.huggingface.co/u/pranavdhoolia)\ **Replies:** 2\ **Last updated:** [October 9, 2026, 1:17am UTC](https://discuss.huggingface.co/t/the-text-we-compare-with-the-first-message-changed-the-cluster-match-from-21-to-92-how-do-we-find-which-change-mattered/190123 "2026-10-09T01:17:04Z")
I am the founder of Operant Labs, so I have a bias. This is a research run, not a product feature. I want critique of the method. The task: a new agent conversation starts. Match its first user message to one of 4 known… --- ## [arXiv Endorsement Request for cs.CR / cs.AI](https://discuss.huggingface.co/t/arxiv-endorsement-request-for-cs-cr-cs-ai/178659)
**Author:** [@Fleed17](https://discuss.huggingface.co/u/Fleed17)\ **Replies:** 4\ **Last updated:** [October 8, 2026, 8:27am UTC](https://discuss.huggingface.co/t/arxiv-endorsement-request-for-cs-cr-cs-ai/178659 "2026-10-08T08:27:43Z")
Hi everyone, I’m preparing to submit my first paper to arXiv in the cs.CR (Cryptography and Security) and cs.AI (Artificial Intelligence) categories, and I need an endorsement from an established arXiv author in either … [Next page](https://discuss.huggingface.co/c/research/7.md?page=1)