Daily Issue
Vol. I — No. 14
14 · 05
Thursday, 14 May 2026
Generated 2026-05-14 11:28
google/gemini-2.5-flash-lite-preview-09-2025
我是一个经常笑的人，可我不是经常开心的人。 — 未闻花名 44 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Overcast
Today's range
13.1°↓7.2°
currently 11.2°
Feels
8.3°
Rain
94%
Wind
10 km/h
Humid
61%
Rise
05:09
Set
20:43
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-05-14
0 Bullish
0 Bearish
12 Neutral
Sector Tape
Servers and Thermal Management 2 names
64 Top: VRT · Neutral · RS -1.5% Bullish 0 / Bearish 0 / 5d +2.6%
Foundry 2 names
62 Top: TSM · Neutral · RS -3.3% Bullish 0 / Bearish 0 / 5d +0.9%
Networking Equipment 4 names
60 Top: CIEN · Neutral · RS -8.7% Bullish 0 / Bearish 0 / 5d -4.7%
Battery and Energy Storage 3 names
59 Top: FLNC · Neutral · RS +26.8% Bullish 0 / Bearish 0 / 5d +27.0%
Hyperscale Cloud 4 names
59 Top: GOOGL · Neutral · RS -3.8% Bullish 0 / Bearish 0 / 5d -1.2%
Manufacturing 4 names
59 Top: FLEX · Neutral · RS -1.3% Bullish 0 / Bearish 0 / 5d -0.2%
Energy Infrastructure 1 names
46 Top: VST · Neutral · RS -9.3% Bullish 0 / Bearish 0 / 5d -9.9%
Compute Mining 4 names
51 Top: CIFR · Neutral · RS -2.7% Bullish 0 / Bearish 0 / 5d -5.8%
Ticker Setup Move Score Evidence Quality
GOOGL Alphabet-A Hyperscale Cloud
Neutral Sector tailwind Low confidence
-0.2% $401.68 5d +1.1%
63 sector positive RS -1.4%

Watchlist item from positive sector tape, 3 recent headline(s).

How SpaceX Helped Send Alphabet Stock to the Moon. - Moomoo Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 3
INTC Intel Foundry
Neutral Sector tailwind Low confidence
-3.6% $115.91 5d +6.4%
61 sector positive RS +2.3%

Watchlist item from -3.6% vs previous close, positive sector tape, 3 recent headline(s).

How Intel Stock Gained 430% - Trefis Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 3
quotes: nasdaq 24 24/24news: google_news_rss 24 24/24filings: sec 24 24/24, fallback 24

Generated from public market data and news for research and education. Not financial advice; data may be delayed, incomplete, or wrong.

§ II

From the arXiv

arXiv preprints 10 of 20
cs.AIarxiv:2605.13652v1Lead article

Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training

Namrata Shivagunde, Vijeta Deshpande, Sherin Muckatira, Anna Rumshisky

his paper moves beyond simple perplexity comparisons to geometrically and spectrally analyze the solutions produced by five distinct low-rank pre-training methods against full-rank training. The core contribution is a rigorous characterization of how rank constraints alter the learned internal representations and loss landscape positions, addressing whether low-rank models generalize comparably to their full-rank counterparts.

Read abstract →Full PDFLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.13825v1

History Anchors: How Prior Behavior Steers LLM Decisions Toward Unsafe Actions

Alberto G. Rodríguez Salgado

This paper introduces **HistoryAnchor-100**, a benchmark to test if prior harmful actions steer Large Language Models (LLMs) toward continued unsafe behavior. The core finding is that frontier LLMs, even highly aligned ones, exhibit a striking vulnerability: a…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.13625v1

How to Interpret Agent Behavior

Jie Gao, Kaiser Sun et al.

This paper introduces **ACT*ONOMY**, a novel, three-level hierarchical taxonomy (10 actions, 46 subactions, 120 leaf categories) designed to systematically describe and analyze the runtime behavior of autonomous agents from their natural-language traces. The c…

abstractpdfLINNET · SCORE9RELEVANCE
Why do we need Act· onomy ? Act· onomy can be used to label agent trajectories with human-readable action tags; we use a 13-turn SWE-bench trajectory as a running example. Top: A phase overview of the trajectory on pylint-dev/pylint-5859 , with color-coded regions marking distinct turns. Middle: Three pivotal turns annotated with Act· onomy sub-action tags: Turn 4 ( confirm ) verifies the bug and pivots to code localization; Turn 6 ( stumble ) detects a failed fix and recovers with a new search strategy; Turn 9 ( pinpoint ) identifies \( \b \) in the regex as the root cause. Bottom: A sentence-level zoom into Turn 9, grounding each tag in a specific quoted span from the agent’s Observation → \( \to \) Thought → \( \to \) Action loop.
Why do we need Act· onomy ? Act· onomy can be used to label agent trajectories with human-readable action tags; we use a 13-turn SWE-bench trajectory as a running example. Top: A phase overview of the…
Task-Centric Taxonomy of Blind Assistance and Distribution of Assistive Task Instances. Distribution of 778 assistive task instances across four domains and their subcategories, highlighting dominant needs in Reading and Text Access (35%) and Mobility and Safety (34%).
Task-Centric Taxonomy of Blind Assistance and Distribution of Assistive Task Instances. Distribution of 778 assistive task instances across four domains and their subcategories, highlighting dominant …
cs.AIarxiv:2605.13579v1

Position: Assistive Agents Need Accessibility Alignment

Jie Hu, Changyuan Yan et al.

This paper argues that current assistive AI systems fail BVI users because they are designed assuming sighted interaction and low-cost verification. The core contribution is introducing the concept of **accessibility alignment** as a first-class design objecti…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.13737v1

Senses Wide Shut: A Representation-Action Gap in Omnimodal LLMs

Trung Nguyen Quang, Yiming Gao et al.

This paper introduces IMAVB, a benchmark to test if omnimodal LLMs can detect contradictions between a textual premise and their own sensory input (vision/audio). The core finding is a "Representation-Action Gap": models reliably encode these premise-perceptio…

abstractpdfLINNET · SCORE9RELEVANCE
Overview of the Representation–Action Gap on IMAVB.
Overview of the Representation–Action Gap on IMAVB.
№06
cs.AI
9

Temper and Tilt Lead to SLOP: Reward Hacking Mitigation with Inference-Time Alignment

Ye Wang, Jing Liu et al.

This paper introduces **SLOP (Sharpened Logarithmic Opinion Pool)**, an extension of inference-time alignment that generalizes techniques to combine ensembles of generative reward …

abstractpdf
№07
cs.AI
9

Where Does Reasoning Break? Step-Level Hallucination Detection via Hidden-State Transport Geometry

Tyler Alvarez, Ali Baheri

This paper introduces a novel method for detecting step-level hallucinations in LLM reasoning by analyzing the geometry of the hidden-state trajectory during a single forward pass.…

abstractpdf
№08
cs.CL
9

Good Agentic Friends Do Not Just Give Verbal Advice: They Can Update Your Weights

Wenrui Bao, Huan Wang et al.

This paper introduces TFlow, a novel weight-space communication framework for multi-agent LLMs that replaces costly natural language message passing with direct weight updates. The…

abstractpdf
№09
cs.AI
8

AttenA+: Rectifying Action Inequality in Robotic Foundation Models

Daojie Peng, Fulong Ma et al.

This paper introduces **AttenA+**, a framework designed to address the "action inequality" in robotic foundation models where all actions are treated equally during training. Atten…

abstractpdf
№10
cs.AI
8

Children's English Reading Story Generation via Supervised Fine-Tuning of Compact LLMs with Controllable Difficulty and Safety

Qian Shen, Fanghua Cao et al.

This paper introduces a method for generating controllable and age-appropriate children's English reading stories by **supervised fine-tuning compact (8B-parameter) LLMs** using ex…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 3
264
pts
Top story
Tell HN: Dont use Claude Design, lost access to my projects after unsubscribing

A user reported losing access to their projects after unsubscribing from Claude Design, warning others about potential data loss with the service.

13 May discuss on HN →
205
The US is winning the AI race where it matters most: commercialization
avkcode.github.io13 May
169
Meta won't let you block its AI account on Threads
theverge.com13 May
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 14 · thank you for reading