Daily Issue
Vol. I — No. 12
11 · 05
Monday, 11 May 2026
Generated 2026-05-11 12:33
google/gemini-2.5-flash-lite-preview-09-2025
君子坦荡荡，小人长戚戚。 — 论语 46 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Light drizzle
Today's range
12.0°↓5.1°
currently 10.2°
Feels
7.1°
Rain
73%
Wind
12 km/h
Humid
67%
Rise
05:14
Set
20:39
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-05-11
6 Bullish
0 Bearish
6 Neutral
Sector Tape
Servers and Thermal Management 2 names
73 Top: VRT · Bullish · RS +4.0% Bullish 2 / Bearish 0 / 5d +13.7%
Foundry 2 names
72 Top: INTC · Bullish · RS +3.0% Bullish 1 / Bearish 0 / 5d +14.5%
Compute Mining 4 names
68 Top: WULF · Neutral · RS +2.7% Bullish 1 / Bearish 0 / 5d +23.1%
Hyperscale Cloud 4 names
66 Top: ORCL · Neutral · RS -1.4% Bullish 1 / Bearish 0 / 5d +4.9%
Manufacturing 4 names
64 Top: FLEX · Neutral · RS +5.3% Bullish 1 / Bearish 0 / 5d +11.0%
Energy Infrastructure 1 names
39 Top: VST · Neutral · RS -0.2% Bullish 0 / Bearish 0 / 5d -4.9%
Networking Equipment 4 names
61 Top: CIEN · Neutral · RS -14.3% Bullish 0 / Bearish 0 / 5d -5.8%
Battery and Energy Storage 3 names
53 Top: FLNC · Neutral · RS +38.9% Bullish 0 / Bearish 0 / 5d +37.4%
Ticker Setup Move Score Evidence Quality
INTC Intel Foundry
Bullish Gap up + news High confidence
+4.2% $130.17 5d +25.4%
76 sector positive RS +14.0%

Bullish setup from +4.2% vs previous close, positive sector tape, 3 recent headline(s).

INTC Stock Quote Price and Forecast - CNN Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
DELL Dell Technologies Servers and Thermal Management
Bullish Sector tailwind Medium confidence
-1.6% $256.18 5d +23.9%
71 sector positive RS +14.2%

Bullish setup from -1.6% vs previous close, positive sector tape, 3 recent headline(s).

How Dell Technologies Stock Gained 120% - Trefis Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
HUT Hut 8 Compute Mining
Neutral Sector tailwind Medium confidence
-1.3% $97.20 5d +27.9%
69 sector positive RS +7.5%

Watchlist item from -1.3% vs previous close, positive sector tape, 3 recent headline(s).

Northland Securities Has Weak Forecast for Hut 8 Q2 Earnings - MarketBeat Needs fresh price/news confirmation before becoming an actionable setup.
quote: delayed fallback news: fresh financials: fresh news: 3
GOOGL Alphabet-A Hyperscale Cloud
Neutral Sector tailwind Medium confidence
-1.1% $396.55 5d +3.9%
68 sector positive RS -2.5%

Watchlist item from -1.1% vs previous close, positive sector tape, 3 recent headline(s).

Is Alphabet a buy amid soaring Q1 profits on AI cloud growth? - MSN 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 22 22/24filings: 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:2605.07926v1Lead article

AgentEscapeBench: Evaluating Out-of-Domain Tool-Grounded Reasoning in LLM Agents

Zhengkang Guo, Yiyang Li, Lin Qiu, Xiaohua Wang, Jingwen Xv

gentEscapeBench is a novel benchmark designed to evaluate LLM agents' ability to perform complex, out-of-domain tool-grounded reasoning. It uses escape-room style tasks with long-range dependencies, requiring agents to infer and execute multi-step procedures involving real external tools and state tracking. The benchmark reveals a significant performance drop for both models and humans as the dependency depth increases, highlighting a critical challenge in agent robustness.

Read abstract →Full PDFLINNET · SCORE9RELEVANCE
Conceptual illustration of AgentEscapeBench. The agent is placed in a themed escape room populated with unfamiliar tools and hidden items. It must explore the environment, invoke tools with correct parameters derived from narrative clues, and propagate intermediate outputs through a multi-step dependency chain to unlock the final exit.
Conceptual illustration of AgentEscapeBench. The agent is placed in a themed escape room populated with unfamiliar tools and hidden items. It must explore the environment, invoke tools with correct parameters derived from narrative clues, and propagate intermediate outputs throug…
GraphDPO pipeline for LLM alignment. For each prompt, the policy samples K K rollouts, which are grouped into equivalence classes according to preference signals. These classes induce a DAG structure whose edges encode dominance relations between groups, with an optional ground-truth node as a global anchor. Equivalence-class masking removes intra-group comparisons so that each response is contrasted only with strictly worse groups via a local Plackett–Luce loss. The resulting losses are aggregated over the graph to update the policy while enforcing transitive preference structure.
GraphDPO pipeline for LLM alignment. For each prompt, the policy samples K K rollouts, which are grouped into equivalence classes according to preference signals. These classes induce a DAG structure …
cs.AIarxiv:2605.08037v1

Beyond Pairs: Your Language Model is Secretly Optimizing a Preference Graph

Ning Liu, Chuanneng Sun et al.

This paper introduces **Graph Direct Preference Optimization (GraphDPO)**, a principled generalization of DPO that moves beyond simple pairwise comparisons. GraphDPO leverages richer preference data structured as directed acyclic graphs (induced by ranked roll…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.07830v1

CyBiasBench: Benchmarking Bias in LLM Agents for Cyber-Attack Scenarios

Taein Lim, Seongyong Ju et al.

This paper introduces **CyBiasBench**, a comprehensive benchmark to quantify the attack-selection bias exhibited by LLM agents in cyber-attack scenarios. The core method involves systematically testing five agents across various targets and prompts to reveal t…

abstractpdfLINNET · SCORE9RELEVANCE
Attack-Selection Bias of LLM Agents. To illustrate attack-selection bias, we measure per-agent average selection rates across the bias observation setting (solid line) and compare them with the corresponding attack success rates (dashed line). The results reveal clear biases in agent behavior.
Attack-Selection Bias of LLM Agents. To illustrate attack-selection bias, we measure per-agent average selection rates across the bias observation setting (solid line) and compare them with the corres…
VGDL game paradigm. (A) Games are defined by combining game rules with map layouts to produce interactive environments. (B) Example Trial Structure of VGDL-fMRI Dataset. Color denotes game names: ( Bait , Chase , Helper , Lemmings , Plaque Attack , Zelda ). All participants played the same level progression structure with randomized game order. The subsequent levels reveal new rules incrementally. The Interactive Catalogue A lets readers try each game in the browser and browse all participant and LRM agent gameplay replays. Project page: https://botcs.github.io/reason-to-play/
VGDL game paradigm. (A) Games are defined by combining game rules with map layouts to produce interactive environments. (B) Example Trial Structure of VGDL-fMRI Dataset. Color denotes game names: ( Ba…
cs.AIarxiv:2605.08019v1

Reason to Play: Behavioral and Brain Alignment Between Frontier LRMs and Human Game Learners

Botos Csaba, Sreejan Kumar et al.

This paper investigates whether frontier Large Reasoning Models (LRMs) can mimic human learning and planning in novel game environments. The core method involves jointly evaluating LRMs against RL agents using human gameplay data, concurrent fMRI recordings, a…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.08060v1

The Memory Curse: How Expanded Recall Erodes Cooperative Intent in LLM Agents

Jiayuan Liu, Tianqin Li et al.

This paper introduces the "memory curse," demonstrating that expanding the context window for LLM agents systematically *erodes* cooperation in multi-agent social dilemmas. The core mechanism identified is not increased paranoia, but the degradation of forward…

abstractpdfLINNET · SCORE9RELEVANCE
Schematic of repeated social dilemma interactions between two LLM agents with shared memory.
Schematic of repeated social dilemma interactions between two LLM agents with shared memory.
№06
cs.AI
9

Tool Calling is Linearly Readable and Steerable in Language Models

Zekun Wu, Ze Wang et al.

This paper demonstrates that the tool selection within language models is **linearly readable and steerable** by analyzing internal activations across various models. By manipulati…

abstractpdf
№07
cs.LG
9

RelAgent: LLM Agents as Data Scientists for Relational Learning

Xingyue Huang, Louis Tichelman et al.

RelAgent is an LLM-based autonomous agent designed for relational learning, operating in two phases. First, the agent uses tools to autonomously construct feature-generating SQL pr…

abstractpdf
№08
cs.LG
9

Self-Play Enhancement via Advantage-Weighted Refinement in Online Federated LLM Fine-Tuning with Real-Time Feedback

Seohyun Lee, Wenzhi Fang et al.

This paper introduces SPEAR (Self-Play Enhancement via Advantage-Weighted Refinement), an efficient online learning algorithm for federated LLM fine-tuning. SPEAR enables a self-im…

abstractpdf
№09
cs.CL
9

Beyond "I cannot fulfill this request": Alleviating Rigid Rejection in LLMs via Label Enhancement

Ying Zhang, Congyu Qiao et al.

This paper introduces **LANCE** to combat rigid rejection in LLMs by moving beyond binary refusal. LANCE uses variational inference to enhance safety labels, predicting a continuou…

abstractpdf
№10
cs.CL
9

GLiGuard: Schema-Conditioned Classification for LLM Safeguard

Urchade Zaratiana, Mary Newhauser et al.

GLiGuard reframes LLM content moderation as a schema-conditioned classification task, moving away from slow, large autoregressive models. It uses a small (0.3B parameter) bidirecti…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 5
1302
pts
Top story
Local AI needs to be the norm

The core message advocates for local AI processing to become the standard practice rather than relying solely on cloud-based solutions.

unix.foo10 May discuss on HN →
270
Maryland citizens hit with $2B power grid upgrade for out-of-state AI
tomshardware.com10 May
210
An AI coding agent, used to write code, needs to reduce your maintenance costs
jamesshore.com10 May
167
PS3 Emulator Devs Politely Ask That People Stop Flooding It with AI PRs
kotaku.com10 May
110
Chrome's AI features may be hogging 4GB of your computer storage
theverge.com10 May
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 12 · thank you for reading