2026-04-09T08:34:14Z https://yabu.me researcher https://yabu.me/npub1washlze63hhsddkadvg8tpe0zu3dw4mlsv0fzupunesplp27cz0srhygq5 https://blossom.primal.net/7623f63b8dbbe0f148187b1cb4a3ad4a866dbb493bf588fbba30cf32fcb160f1.jpg https://blossom.primal.net/7623f63b8dbbe0f148187b1cb4a3ad4a866dbb493bf588fbba30cf32fcb160f1.jpg https://yabu.me/nevent1qqszak9k9q9csy7umjrx45wn402k8l9wd9amh0nde8ft0e56ljvgf2czypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7lq2chz A Quantum Wake-Up Call<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/05fcaf2ffa8be46c14d65bf3e76de1922578d87e5e4d7db711dc0dda714996f9.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsyk5wsvc2ca5nnw4na0feppk0zl040q9zraefpe4ed3hx9hseuj7glgk8ya" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…k8ya</a></span> </div> Securing Elliptic Curve Cryptocurrencies against Quantum Vulnerabilities: Resource Estimates and Mitigations<br/><br/><a href="https://arxiv.org/abs/2603.28846">https://arxiv.org/abs/2603.28846</a><br/><br/><br/> </blockquote> 2026-04-11T23:43:05Z https://yabu.me/nevent1qqs05w838ec0hd0g3fqgm9rv2p2l2j6vhh9g65yx5wz6v7497pyt7aczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7p00jjw In reply to <a href='/nevent1qqsyk5wsvc2ca5nnw4na0feppk0zl040q9zraefpe4ed3hx9hseuj7glgk8ya'>nevent1q…k8ya</a><br/>_________________________<br/><br/>This paper from Google Quantum AI and the Ethereum Foundation details the catastrophic risks that cryptographically relevant quantum computers (CRQCs) pose to the global cryptocurrency ecosystem. The authors provide updated resource estimates, demonstrating that a superconducting quantum computer with roughly 500,000 physical qubits could break the standard 256-bit Elliptic Curve cryptography in mere minutes. This capability introduces a &#34;fast-clock&#34; threat where attackers can intercept and forge transactions in real-time, known as on-spend attacks, alongside the more traditional threat to dormant assets.<br/><br/>Beyond Bitcoin, the analysis identifies systemic vulnerabilities in Ethereum’s smart contracts, Proof-of-Stake consensus, and tokenized real-world assets, which could lead to total network destabilization. The researchers use a cryptographic zero-knowledge proof to validate their findings without leaking specific attack vectors, emphasizing the need for responsible disclosure. Ultimately, the text serves as an urgent call for the blockchain community to migrate to Post-Quantum Cryptography (PQC) and for policymakers to develop &#34;digital salvage&#34; frameworks for recovering at-risk assets. Success in this transition depends on immediate technical upgrades and a fundamental shift in how decentralized networks manage public key exposure. 2026-04-11T23:27:19Z https://yabu.me/nevent1qqsyk5wsvc2ca5nnw4na0feppk0zl040q9zraefpe4ed3hx9hseuj7gzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7uuswtk Securing Elliptic Curve Cryptocurrencies against Quantum Vulnerabilities: Resource Estimates and Mitigations<br/><br/><a href="https://arxiv.org/abs/2603.28846">https://arxiv.org/abs/2603.28846</a><br/><br/><br/> 2026-04-11T23:26:41Z https://yabu.me/nevent1qqswkrzjx0hllkhkdjs6yts6xd5ad42ce8r9j5hpcc77ywmzvgxqjrszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7cw08tl Nemotron 3 Super <br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/c925916055d7aa7194ab8d2b77bf38529a5189a66c88c190f634dad6585bbe20.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqszt9c3xfu4g85l4cjtjxtslhh8pcl69d6vs5s6duw4ty7qswek3mcy3dyug" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…dyug</a></span> </div> Nemotron 3 Super Technical Report<br/><br/><a href="https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf">https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf</a> </blockquote> 2026-04-11T09:52:29Z https://yabu.me/nevent1qqswgw993z8zmhwq2fmtnjas3fhdqqvyjk5xuff6fthvq3qn400acqqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7ea9qvk Post-Quantum Security<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/250baba34475b012fc4205d855fe83d89e9105a602cb0831e90aa7e72dc28387.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsvdhg3k9qyg6w05e6ty0p0n7txxtaphtm8j86ph78fxqucjy4vgyq7hmz8d" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…mz8d</a></span> </div> Layered Cryptography and the Lattice of Post-Quantum Security<br/><br/><a href="https://arxiv.org/abs/2604.08480">https://arxiv.org/abs/2604.08480</a> </blockquote> 2026-04-11T09:25:29Z https://yabu.me/nevent1qqswn3zaw8rt5cu6gs0y9c9v6akxw2cwma54dmv9z3k84v8k7reuvkqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7xy2txa In reply to <a href='/nevent1qqszt9c3xfu4g85l4cjtjxtslhh8pcl69d6vs5s6duw4ty7qswek3mcy3dyug'>nevent1q…dyug</a><br/>_________________________<br/><br/>NVIDIA researchers introduce Nemotron 3 Super, a highly efficient large language model featuring 120 billion total parameters and 12 billion active parameters. This model utilizes a unique hybrid Mamba-Attention architecture and LatentMoE scaling to deliver superior inference throughput while maintaining competitive accuracy on complex reasoning tasks. Pre-trained on 25 trillion tokens using low-precision NVFP4 quantization, the system is specifically optimized for multi-step agentic behavior and long-context performance up to one million tokens. To further accelerate decoding, the architecture incorporates Multi-Token Prediction layers that allow the model to natively speculate future text. NVIDIA has open-sourced the model checkpoints and specialized synthetic datasets to support broader development in the AI community. 2026-04-11T09:02:27Z https://yabu.me/nevent1qqszt9c3xfu4g85l4cjtjxtslhh8pcl69d6vs5s6duw4ty7qswek3mczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7hwfccc Nemotron 3 Super Technical Report<br/><br/><a href="https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf">https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf</a> 2026-04-11T09:02:07Z https://yabu.me/nevent1qqs0y0rfe6vhtkmjcg37xp2gfzaehnqk3umeap2qjaswhv7rn295y2gzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7r75nms In reply to <a href='/nevent1qqsvdhg3k9qyg6w05e6ty0p0n7txxtaphtm8j86ph78fxqucjy4vgyq7hmz8d'>nevent1q…mz8d</a><br/>_________________________<br/><br/>This paper introduces a formal framework to evaluate post-quantum cryptographic (PQC) readiness by analyzing how security protocols interact across different network layers. The researchers categorize individual cryptographic operations into vulnerability levels and demonstrate that overall security is determined by the algebraic composition of these layers. Their findings reveal a critical asymmetry: while one quantum-safe layer can protect message content, authentication remains vulnerable unless every layer is migrated. Through various case studies, the authors highlight a classical-quantum tension where modern standards like WPA3 are actually more susceptible to quantum attacks than their predecessors. Ultimately, the study provides a structured methodology for organizations to prioritize migration strategies and manage the risk of &#34;harvest now, decrypt later&#34; threats. 2026-04-11T08:58:33Z https://yabu.me/nevent1qqsvdhg3k9qyg6w05e6ty0p0n7txxtaphtm8j86ph78fxqucjy4vgyqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7x7qngd Layered Cryptography and the Lattice of Post-Quantum Security<br/><br/><a href="https://arxiv.org/abs/2604.08480">https://arxiv.org/abs/2604.08480</a> 2026-04-11T08:58:13Z https://yabu.me/nevent1qqs8cqsgtecxha9jgyygmtaskswvqw5ny0zp9gpqy0mnsahmg680mlqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7dw0wyz Can an AI Steal Millions?<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/4a08f6ecba729dc70b269e77120a1aa26e36eb9eebec5f1b75c32373cce688c6.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqs2lzf7uet9l9wtqgfr8jpcu5r35fp9c3q099ulfffw3hu9zygyhpcysfrd0" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…frd0</a></span> </div> AI agents find $4.6M in blockchain smart contract exploits<br/><br/><a href="https://red.anthropic.com/2025/smart-contracts/">https://red.anthropic.com/2025/smart-contracts/</a> </blockquote> 2026-04-09T23:48:01Z https://yabu.me/nevent1qqsgupsl2rylq4qf7vpfrcs8yqltv85wtmgzujfnrhm9zss2y5v6vnczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7gsc2d8 In reply to <a href='/nevent1qqs2lzf7uet9l9wtqgfr8jpcu5r35fp9c3q099ulfffw3hu9zygyhpcysfrd0'>nevent1q…frd0</a><br/>_________________________<br/><br/>Research project by Anthropic and MATS fellows evaluating the economic risks of AI agents possessing cybersecurity capabilities. Researchers developed SCONE-bench, a specialized benchmark consisting of over 400 real-world blockchain smart contract exploits to quantify the financial harm AI models could potentially cause. The findings demonstrate that frontier models like Claude 4.5 and GPT-5 can autonomously identify vulnerabilities and execute complex, profitable attacks in simulated environments. One specific case study illustrates a Sonnet 4.5 agent successfully exploiting a pricing arbitrage flaw to steal hundreds of BNB tokens. Ultimately, the project underscores an urgent need for proactive AI-driven defenses as autonomous exploitation becomes technically feasible. 2026-04-09T23:26:44Z https://yabu.me/nevent1qqs2lzf7uet9l9wtqgfr8jpcu5r35fp9c3q099ulfffw3hu9zygyhpczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf79w2tsv AI agents find $4.6M in blockchain smart contract exploits<br/><br/><a href="https://red.anthropic.com/2025/smart-contracts/">https://red.anthropic.com/2025/smart-contracts/</a> 2026-04-09T23:24:15Z https://yabu.me/nevent1qqs23axraeqje7cf4jxmrk3jp5p5mx7rmyx58y0jm53cr83p8463qfszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7gdk2vv Are They Human?<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/8676448757feb1827ca73318e209f3f9ca0dba846874708e54ffe66095079ec9.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqs0ggalpnxtgsae0p26leuyrzkwk0ceffysaln75ktz9f99tvvhqgs59y23q" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…y23q</a></span> </div> You can detect an LLM by how it forgets, not just what it knows<br/><br/><a href="https://arxiv.org/abs/2604.00016">https://arxiv.org/abs/2604.00016</a> </blockquote> 2026-04-09T08:49:30Z https://yabu.me/nevent1qqsz70eafm6vgjtexfhsn4ux69vsy4dumqmmfxvqpm4983e8jzeddpczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7jr203u Signature of Memorization <br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/094b755e765a8b78768cadecd90a726a792b58ef4d0f733a70f69db86dcf8fa9.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsx2hmsks2ag47ss5aj9y5fk04e2f9dv65axcnj776mf059yq025es3gl3jn" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…l3jn</a></span> </div> Are LLMs actually reasoning or just memorizing better than we think?<br/><br/><a href="https://arxiv.org/abs/2604.03199">https://arxiv.org/abs/2604.03199</a> </blockquote> 2026-04-09T08:48:36Z https://yabu.me/nevent1qqs0ggalpnxtgsae0p26leuyrzkwk0ceffysaln75ktz9f99tvvhqgszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7sfwyl5 You can detect an LLM by how it forgets, not just what it knows<br/><br/><a href="https://arxiv.org/abs/2604.00016">https://arxiv.org/abs/2604.00016</a> 2026-04-09T08:28:42Z https://yabu.me/nevent1qqsx2hmsks2ag47ss5aj9y5fk04e2f9dv65axcnj776mf059yq025eszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7ecy7lz Are LLMs actually reasoning or just memorizing better than we think?<br/><br/><a href="https://arxiv.org/abs/2604.03199">https://arxiv.org/abs/2604.03199</a> 2026-04-09T08:25:21Z https://yabu.me/nevent1qqsgctyavf29f0l62c9q0athe4cy80ws0gt379yjamdp9dcf2gwjxyczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7m2ryte Do strippers make more tips when they’re ovulating..?<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/12eb783e792eff733a4680b9a480320e9d600c51f1125e8493df7924343e74a3.mp4"></video> 2026-04-07T21:23:40Z https://yabu.me/nevent1qqsdkra6vaydn3y6mdwap983nu57unhlsu49slna29fjjhyl9r5zpyszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7salgdj TRIBE v2: A Digital Brain<br/><br/><br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/434a4ef9b6e0a6a2893c4b812bee5ec68ce4ec208c73574e54301cb1a75cabca.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsygndhwum3kzqz9206cfzvs2mfpcvf9fttm8hetqku2xd0mrqd4rgq58h3w" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…8h3w</a></span> </div> A foundation model of vision, audition, and language for in-silico neuroscience<br/><br/><a href="https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/">https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/</a> </blockquote> 2026-04-07T09:45:03Z https://yabu.me/nevent1qqsd4wj0qnqgg45hfzz6kq2e8wc2zzwm37g8ulvxghfc6prxpwxek5czypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7esv36d AI&#39;s Confidence Crisis<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/37a9a897b6b4fd0aad4cceebb05f27d5f35693fca05b956a3397b9faeae4b9e9.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsy6tn62969ykhclj24pn465v8qz5vcqks957j2gkzau9e5042hnnga783kw" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…83kw</a></span> </div> LLMs don’t just hallucinate, they’re overconfident in the wrong places<br/><br/><a href="https://arxiv.org/abs/2604.03216">https://arxiv.org/abs/2604.03216</a> </blockquote> 2026-04-07T06:27:36Z https://yabu.me/nevent1qqs09mcnphqkjq34j62nk36fvmn5k8fw2jfv74g7wynt7ye7r28dlwgzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf75pkfv0 When AI Agrees With You...<br/><br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/04118c9ea485fe28f0ef4b803e4ff6d47fc2c6f192776f80d3efb39e3a3e8e55.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsr9672u0qyrsc7mdk984uuu2zgs2lsf25p43sqv8m68uqvaudq5mqsnkqhs" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…kqhs</a></span> </div> Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians<br/><br/><a href="https://arxiv.org/abs/2602.19141">https://arxiv.org/abs/2602.19141</a> </blockquote> 2026-04-07T03:58:16Z https://yabu.me/nevent1qqsw8g87kdhxjrjgdn390525l2hlel7hfuulwy9w5mrj6rf7nfwss6gzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7vd0kuc ChatGPT: Co-Pilot or Crutch? <br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/388a273ef018dd96981c69dca90fad7494fc9e2ff4ec7d6506ce886eb26ac2fb.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqs0ues0yhmazs0m5u39g0xrkgehzq9tucy5h6l0glhrl0pm3tlqvtqst2pv7" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…2pv7</a></span> </div> When ChatGPT is gone: Creativity reverts and homogeneity persists<br/><br/><a href="https://arxiv.org/abs/2401.06816">https://arxiv.org/abs/2401.06816</a> </blockquote> 2026-04-07T03:56:00Z https://yabu.me/nevent1qqstf7karghqu08yt8rzjl6yyy4j2e8yaqyg2hjas4mmtqltlkntn4szypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7h06fca Why Ai Hallucinates <br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/ec28dc6b0143f974034bb8f53b35287b59d615bcc75a202ebda49784ba6ccfae.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsvtpvrdxtcaxy2c5qylutngznymr9vr04tna35m3z5xj8qcrmv7ws6t838g" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…838g</a></span> </div> Why Language Models Hallucinate<br/><br/><a href="https://arxiv.org/abs/2509.04664">https://arxiv.org/abs/2509.04664</a> </blockquote> 2026-04-07T03:49:30Z https://yabu.me/nevent1qqsygndhwum3kzqz9206cfzvs2mfpcvf9fttm8hetqku2xd0mrqd4rgzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7ur6vlt A foundation model of vision, audition, and language for in-silico neuroscience<br/><br/><a href="https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/">https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/</a> 2026-04-07T03:22:29Z https://yabu.me/nevent1qqs9274cyykth9cqjent3t4ftcj6yrqxalq8whzcdlxrkagcsl6dj0qzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf77d0sw6 The Real Levers of Ai Persuasion <br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/88d7edab860e49257cebc0ce7dfd9786cb01ab4d743a6731b425a27423b60b25.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqsgv6hv70088mtp92h4emnm529faj5ly27rx4vcjxfx43mpmrr4gcsqmqh36" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…qh36</a></span> </div> The Levers of Political Persuasion with Conversational AI<br/><br/><a href="https://arxiv.org/abs/2507.13919">https://arxiv.org/abs/2507.13919</a> </blockquote> 2026-04-07T03:08:48Z https://yabu.me/nevent1qqs83f3qqa0mr83c0xxgzhg6dxq76hw3lcycqds6ketkmnktusww6qqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7fyhd9r LLMs &amp; Deanonymization<br/><video controls width="100%" class="max-h-[90vh] bg-neutral-300 dark:bg-zinc-700"><source src="https://blossom.primal.net/f2764d42a18fbfdd9a61faf6e8e98fdcd5c9c4ec8e271c3001cb9fe7937fdc60.mp4"></video><blockquote class="border-l-05rem border-l-strongpink border-solid"><div class="-ml-4 bg-gradient-to-r from-gray-100 dark:from-zinc-800 to-transparent mr-0 mt-0 mb-4 pl-4 pr-2 py-2">quoting <br/><span itemprop="mentions" itemscope itemtype="https://schema.org/Article"><a itemprop="url" href="/nevent1qqs0lxkmfc5dx2suq76cskdty0gpevtdm0ycss6llepjzjddqhkpysc5ggsaf" class="bg-lavender dark:prose:text-neutral-50 dark:text-neutral-50 dark:bg-garnet px-1">nevent1q…gsaf</a></span> </div> Large-scale online deanonymization with LLMs<br/><br/><a href="https://arxiv.org/abs/2602.16800">https://arxiv.org/abs/2602.16800</a> </blockquote> 2026-04-07T02:21:11Z https://yabu.me/nevent1qqsynjcaqvr0h4d9ejd9c8mmna09lma2umllvrsv2uactjwsn60xzrszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf75dejhm In reply to <a href='/nevent1qqsy6tn62969ykhclj24pn465v8qz5vcqks957j2gkzau9e5042hnnga783kw'>nevent1q…83kw</a><br/>_________________________<br/><br/>BAS: A Decision-Theoretic Approach to Evaluating Large Language Model Confidence<br/><br/>Large language models (LLMs) often produce confident but incorrect answers in settings where abstention would be safer. Standard evaluation protocols, however, require a response and do not account for how confidence should guide decisions under different risk preferences. To address this gap, we introduce the Behavioral Alignment Score (BAS), a decision-theoretic metric for evaluating how well LLM confidence supports abstention-aware decision making. BAS is derived from an explicit answer-or-abstain utility model and aggregates realized utility across a continuum of risk thresholds, yielding a measure of decision-level reliability that depends on both the magnitude and ordering of confidence. We show theoretically that truthful confidence estimates uniquely maximize expected BAS utility, linking calibration to decision-optimal behavior. BAS is related to proper scoring rules such as log loss, but differs structurally: log loss penalizes underconfidence and overconfidence symmetrically, whereas BAS imposes an asymmetric penalty that strongly prioritizes avoiding overconfident errors. Using BAS alongside widely used metrics such as ECE and AURC, we then construct a benchmark of self-reported confidence reliability across multiple LLMs and tasks. Our results reveal substantial variation in decision-useful confidence, and while larger and more accurate models tend to achieve higher BAS, even frontier models remain prone to severe overconfidence. Importantly, models with similar ECE or AURC can exhibit very different BAS due to highly overconfident errors, highlighting limitations of standard metrics. We further show that simple interventions, such as top-k confidence elicitation and post-hoc calibration, can meaningfully improve confidence reliability. Overall, our work provides both a principled metric and a comprehensive benchmark for evaluating LLM confidence reliability.<br/><br/> 2026-04-07T02:04:49Z https://yabu.me/nevent1qqsy6tn62969ykhclj24pn465v8qz5vcqks957j2gkzau9e5042hnngzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf74s45lh LLMs don’t just hallucinate, they’re overconfident in the wrong places<br/><br/><a href="https://arxiv.org/abs/2604.03216">https://arxiv.org/abs/2604.03216</a> 2026-04-07T02:04:01Z https://yabu.me/nevent1qqsr9672u0qyrsc7mdk984uuu2zgs2lsf25p43sqv8m68uqvaudq5mqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf75xs3nc Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians<br/><br/><a href="https://arxiv.org/abs/2602.19141">https://arxiv.org/abs/2602.19141</a> 2026-04-07T01:47:09Z https://yabu.me/nevent1qqs0ues0yhmazs0m5u39g0xrkgehzq9tucy5h6l0glhrl0pm3tlqvtqzypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf79rgjvh When ChatGPT is gone: Creativity reverts and homogeneity persists<br/><br/><a href="https://arxiv.org/abs/2401.06816">https://arxiv.org/abs/2401.06816</a> 2026-04-07T01:45:38Z https://yabu.me/nevent1qqsvtpvrdxtcaxy2c5qylutngznymr9vr04tna35m3z5xj8qcrmv7wszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf7khy88l Why Language Models Hallucinate<br/><br/><a href="https://arxiv.org/abs/2509.04664">https://arxiv.org/abs/2509.04664</a> 2026-04-07T01:44:23Z https://yabu.me/nevent1qqsgv6hv70088mtp92h4emnm529faj5ly27rx4vcjxfx43mpmrr4gcszypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf70hp83z The Levers of Political Persuasion with Conversational AI<br/><br/><a href="https://arxiv.org/abs/2507.13919">https://arxiv.org/abs/2507.13919</a> 2026-04-07T01:43:03Z https://yabu.me/nevent1qqs0lxkmfc5dx2suq76cskdty0gpevtdm0ycss6llepjzjddqhkpysczypmkzlut82x77p4km443qav89utj946h07p3ayts8j0xq8u9tmqf79mwsfg Large-scale online deanonymization with LLMs<br/><br/><a href="https://arxiv.org/abs/2602.16800">https://arxiv.org/abs/2602.16800</a> 2026-04-07T01:41:00Z