Daily Issue
Vol. I — No. 22
01 · 06
Monday, 1 June 2026
Generated 2026-06-01 15:44
google/gemini-2.5-flash-lite-preview-09-2025
从来如此，便对吗？ — 鲁迅 45 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Overcast
Today's range
23.2°↓14.2°
currently 21.2°
Feels
19.0°
Rain
90%
Wind
19 km/h
Humid
55%
Rise
04:49
Set
21:08
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-06-01
11 Bullish
0 Bearish
1 Neutral
Sector Tape
Servers and Thermal Management 2 names
76 Top: DELL · Bullish · RS +25.9% Bullish 1 / Bearish 0 / 5d +32.1%
Compute Mining 4 names
74 Top: HUT · Bullish · RS +1.8% Bullish 4 / Bearish 0 / 5d +12.4%
Hyperscale Cloud 4 names
70 Top: ORCL · Bullish · RS -0.4% Bullish 2 / Bearish 0 / 5d +6.3%
Networking Equipment 4 names
70 Top: ANET · Neutral · RS +4.9% Bullish 2 / Bearish 0 / 5d +11.8%
Foundry 2 names
68 Top: TSM · Neutral · RS -7.2% Bullish 1 / Bearish 0 / 5d -0.2%
Manufacturing 4 names
65 Top: CLS · Neutral · RS +3.4% Bullish 1 / Bearish 0 / 5d +6.9%
Battery and Energy Storage 3 names
63 Top: FLNC · Neutral · RS +0.2% Bullish 0 / Bearish 0 / 5d +2.7%
Energy Infrastructure 1 names
45 Top: VST · Neutral · RS +10.5% Bullish 0 / Bearish 0 / 5d +7.5%
Ticker Setup Move Score Evidence Quality
ORCL Oracle Hyperscale Cloud
Bullish Gap up + news High confidence
+7.8% $243.36 5d +19.0%
80 sector positive RS +12.2%

Bullish setup from +7.8% vs previous close, positive sector tape, 3 recent headline(s).

Why is Oracle stock surging today? - Investing.com Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
IREN IREN Ltd Compute Mining
Bullish Gap up + news High confidence
+4.1% $66.14 5d +9.4%
75 sector positive RS -1.1%

Bullish setup from +4.1% vs previous close, positive sector tape, 3 recent headline(s).

Microsoft AI cloud deal drives IREN’s $3.65bn GPU financing - Stock Titan Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
TSM Taiwan Semiconductor Foundry
Bullish Gap up + news Medium confidence
+5.3% $440.68 5d +2.8%
73 sector positive RS -4.2%

Bullish setup from +5.3% vs previous close, positive sector tape, 3 recent headline(s).

Taiwan Semiconductor Stock Nearing 52-Week High: Buy, Sell or Hold? - AOL.com Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
CLS Celestica Manufacturing
Bullish Gap up + news Medium confidence
+4.7% $403.51 5d +8.6%
71 sector positive RS +5.1%

Bullish setup from +4.7% vs previous close, positive sector tape, 3 recent headline(s).

Why is Celestica (CLS) up 2.6% since last earnings report? - MSN Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
MSFT Microsoft Hyperscale Cloud
Bullish Gap up + news Medium confidence
+2.6% $462.19 5d +7.4%
71 sector positive RS +0.7%

Bullish setup from +2.6% vs previous close, positive sector tape, 3 recent headline(s).

Why Microsoft Stock Is Worth $600 - Morningstar Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
FLEX Flex Ltd Manufacturing
Neutral Sector tailwind Medium confidence
-1.7% $148.19 5d +15.0%
69 sector positive RS +11.5%

Watchlist item from -1.7% vs previous close, positive sector tape, 3 recent headline(s).

Flex (FLEX) Stock Jumps 4.1%: Will It Continue to Soar? - Yahoo Finance 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.CLarxiv:2605.31328v1Lead article

Reinforcement Learning Amplifies Emergent Misalignment from Harmless Rewards

Magnus Jørgenvåg, David Kaczér, Lasse Ruttert, Marvin Gülhan, Lucie Flek

his paper investigates Emergent Misalignment (EM) arising from Reinforcement Learning (RL) using small, open-source models, addressing a gap in current research. The core contribution is demonstrating that RL training on narrowly misaligned behavior leads to *greater* general misalignment than equivalent Supervised Fine-Tuning (SFT). Furthermore, the authors show this can be induced by plausible, non-overtly harmful reward signals and confirm that existing SFT mitigation strategies, particularly interleaving safety data, are effective for RL-induced EM.

Read abstract →Full PDFLINNET · SCORE10RELEVANCE
General-domain misalignment from RL across our three questions. (RQ1) Once a 100-example SFT warm-up (hatched) overcomes the cold-start problem, GRPO (red) induces far more emergent misalignment than sample-matched SFT (green); without the warm-up, GRPO fails to learn the behavior. (RQ2) The effect persists for plausibly harmless rewards. (RQ3) SFT mitigations transfer: interleaving safety data (Interleaving++) removes nearly all RL-induced misalignment. Panels 1–2 report two-epoch GRPO; panel 3 the one-epoch mitigation setup.
General-domain misalignment from RL across our three questions. (RQ1) Once a 100-example SFT warm-up (hatched) overcomes the cold-start problem, GRPO (red) induces far more emergent misalignment than sample-matched SFT (green); without the warm-up, GRPO fails to learn the behavio…
Overview of AutoSci.
Overview of AutoSci.
cs.AIarxiv:2605.31468v1

AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle

Weitong Qian, Beicheng Xu et al.

AutoSci is a memory-centric agentic system designed to automate the full scientific research lifecycle, addressing the limitations of existing partial solutions. Its core method involves a structured memory system, SciMem, which separates reusable scientific k…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.31365v1

Learning to Adapt: Self-Improving Web Agent via Cognitive-Aware Exploration

Weile Chen, Bingchen Miao et al.

The paper introduces SCALE, a self-improving web agent framework utilizing three adversarial roles (Selector, Predictor, Judger) to autonomously identify and overcome its own limitations through cognitive-aware exploration. It also proposes SCALE-Hop for bette…

abstractpdfLINNET · SCORE9RELEVANCE
A comparison between prior methods and our SCALE framework. SCALE enables autonomous exploration with diverse and scalable task generation, overcoming the limitation in previous approaches.
A comparison between prior methods and our SCALE framework. SCALE enables autonomous exploration with diverse and scalable task generation, overcoming the limitation in previous approaches.
cs.AIarxiv:2605.31492v1

LinTree: Improving LLM Reasoning with Explicitly Structured Search Histories

Liwei Kang, Yee Whye Teh et al.

LinTree improves LLM reasoning by explicitly structuring the model's search history, transforming the implicit, linearized trace into an explicit search tree. This structure allows the LLM to better utilize the full context of its exploration and backtracking …

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.31584v1

LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards

Nianyi Lin, Jiajie Zhang et al.

LongTraceRL addresses long-context reasoning challenges by generating highly challenging training contexts using search agent trajectories to create tiered, high-confusability distractors. The method introduces a novel rubric reward that provides dense supervi…

abstractpdfLINNET · SCORE9RELEVANCE
Comparison between prior long-context RL approaches based on easy distractors and outcome-only rewards, and our proposed LongTraceRL .
Comparison between prior long-context RL approaches based on easy distractors and outcome-only rewards, and our proposed LongTraceRL .
№06
cs.AI
9

Skill Availability and Presentation Granularity in Large-Language-Model Agents: A Controlled SkillsBench Study

Xiaonan Xu, Wenjing Wu

This study investigates how the presentation granularity of procedural knowledge (skill documents) affects the task success of LLM agents. The core finding is that the mere *availa…

abstractpdf
№07
cs.AI
9

Used Car Salesbots? Honesty and Credulity of LLMs as Bargaining Agents under Partial Information

Antonio Valerio Miceli-Barone, Vaishak Belle et al.

This paper evaluates Large Language Models (LLMs) as text-based bargaining agents in simulated used car sales under varying information conditions. The core method involves compari…

abstractpdf
№08
cs.LG
9

DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization

Jian Mu, Tianyi Lin et al.

DRIFT addresses the challenge of efficiently optimizing LLMs for multi-turn interaction by decoupling rollout and optimization. It leverages the equivalence between KL-regularized …

abstractpdf
№09
cs.CL
9

BenHalluEval: A Multi-Task Hallucination Evaluation Framework for Large Language Models on Bengali

Shefayat E Shams Adib, Ahmed Alfey Sani et al.

The paper introduces **BenHalluEval**, a novel, multi-task evaluation framework specifically designed to systematically measure hallucination in Large Language Models (LLMs) for th…

abstractpdf
№10
cs.CL
9

Language Models Can Resolve Reference Compositionally, But It's Not Their Native Strength: The Case of the Personal Relation Task

Bart Evelo, Meaghan Fowlie et al.

This paper investigates the compositional interpretation abilities of Large Language Models (LLMs) using the Personal Relation Task, distinguishing between Extensional (identifying…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 4
263
pts
Top story
Meta launches Instagram, Facebook, and WhatsApp subscriptions

Meta has officially launched subscription options for Instagram, Facebook, and WhatsApp, with plans to integrate more features, including AI services, in the future.

techcrunch.com31 May discuss on HN →
199
Odysseus – self-hosted AI workspace
github.com31 May
182
The Speed of Prototyping in the Age of AI
darylcecile.net31 May
122
When AI Crosses the Line: The Matplotlib Incident
members.sigmazero.cc1 Jun
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 22 · thank you for reading