Daily Issue
Vol. I — No. 9
06 · 05
Wednesday, 6 May 2026
Generated 2026-05-06 11:28
google/gemini-2.5-flash-lite-preview-09-2025
隔着屏幕轻易产生感情的你，肯定很孤独吧。 — you're somebody else 51 items · 4 sections
§ 0

The Morning

Local weather 1
This morning in
London
Overcast
Today's range
13.0°↓8.5°
currently 11.5°
Feels
8.1°
Rain
47%
Wind
14 km/h
Humid
59%
Rise
05:22
Set
20:31
§ I

US Stocks

Pre-market signal radar 12
US pre-market radar
premarket 2026-05-06
12 Bullish
0 Bearish
0 Neutral
Sector Tape
Compute Mining 4 names
80 Top: CIFR · Bullish · RS +0.6% Bullish 4 / Bearish 0 / 5d +18.9%
Servers and Thermal Management 2 names
78 Top: VRT · Bullish · RS +2.8% Bullish 2 / Bearish 0 / 5d +8.4%
Manufacturing 4 names
76 Top: FLEX · Bullish · RS +4.5% Bullish 4 / Bearish 0 / 5d +8.2%
Foundry 2 names
74 Top: INTC · Bullish · RS +6.0% Bullish 1 / Bearish 0 / 5d +14.3%
Networking Equipment 4 names
70 Top: CRDO · Bullish · RS +2.5% Bullish 2 / Bearish 0 / 5d +7.4%
Hyperscale Cloud 4 names
68 Top: ORCL · Neutral · RS +1.8% Bullish 1 / Bearish 0 / 5d +6.0%
Battery and Energy Storage 3 names
64 Top: FLNC · Neutral · RS -5.5% Bullish 0 / Bearish 0 / 5d -1.7%
Energy Infrastructure 1 names
55 Top: VST · Neutral · RS -2.1% Bullish 0 / Bearish 0 / 5d -0.5%
Ticker Setup Move Score Evidence Quality
WULF TeraWulf Compute Mining
Bullish Gap up + news High confidence
+4.6% $24.57 5d +12.9%
80 sector positive RS -5.4%

Bullish setup from +4.6% vs previous close, positive sector tape, 3 recent headline(s).

TeraWulf Inc. $WULF Shares Acquired by Oppenheimer & Co. Inc. - MarketBeat 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
Bullish Gap up + news High confidence
+26.4% $121.89 5d +10.7%
79 sector positive RS +7.0%

Bullish setup from +26.4% vs previous close, positive sector tape, 3 recent headline(s).

Why Is Flex Stock Trending Overnight? - Benzinga 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
Bullish Gap up + news High confidence
+35.4% $109.01 5d +11.7%
79 sector positive RS -6.7%

Bullish setup from +35.4% vs previous close, positive sector tape, 3 recent headline(s).

Hut 8 Reports First Quarter 2026 Results - PR Newswire 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
+5.4% $57.72 5d +23.2%
79 sector positive RS +4.8%

Bullish setup from +5.4% vs previous close, positive sector tape, 3 recent headline(s).

symbol__ 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
VRT Vertiv Holdings Servers and Thermal Management
Bullish Gap up + news High confidence
+4.5% $356.50 5d +11.8%
79 sector positive RS +6.1%

Bullish setup from +4.5% vs previous close, positive sector tape, 3 recent headline(s).

Equitable Trust Co. Takes Position in Vertiv Holdings Co. $VRT - MarketBeat 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
+3.2% $431.09 5d +15.5%
75 sector positive RS +11.9%

Bullish setup from +3.2% vs previous close, positive sector tape, 3 recent headline(s).

Celestica: A Bet On AI CapEx Growth (NYSE:CLS) - Seeking Alpha Weakens if price fades below previous close or sector benchmarks roll over.
quote: delayed fallback news: fresh financials: fresh news: 3
quotes: nasdaq 24 24/24news: google_news_rss 23 23/24filings: 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:2605.03799v1Lead article

Natural Language Processing: A Comprehensive Practical Guide from Tokenisation to RLHF

Mullosharaf K. Arabov

his paper presents a comprehensive, practical practicum guiding users through the entire modern NLP pipeline, from tokenization to RLHF. Its core contribution is providing twelve reproducible research artifacts, requiring public code and model publication for each session, all built around a single evolving corpus. The work emphasizes open-weight models and enriches the material with original research on low-resource languages like Tajik and Tatar.

Read abstract →Full PDFLINNET · SCORE10RELEVANCE
cs.AIarxiv:2605.03900v1

Contextual Multi-Objective Optimization: Rethinking Objectives in Frontier AI Systems

Jie Zhou, Qin Chen et al.

This paper introduces **Contextual Multi-Objective Optimization (CMOO)** to address the unreliability of Frontier AI in open-ended tasks where objectives are ambiguous or context-dependent. The core method involves formulating the problem so that AI systems mu…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.03862v1

Correct Is Not Enough: Training Reasoning Planners with Executor-Grounded Rewards

Tianyang Han, Hengyu Shi et al.

This paper introduces **TraceLift**, a reinforcement learning framework that trains reasoning planners using **executor-grounded rewards**, moving beyond simple final-answer correctness. TraceLift uses a frozen executor to evaluate the utility of the planner's…

abstractpdfLINNET · SCORE9RELEVANCE
The overall framework of TraceLift-Groups and TraceLift . (a) Data curation pipeline of TraceLift-Groups . Then we use TraceLift-Groups to finetune the reward model specialized for reasoning supervising by the designed loss. (b) GRPO training process of the planner using previous trained reasoning reward model. (c) Details of execution calculation process. The Reasoning RM score is weighted by measured executor uplift before being combined with verifier feedback for planner optimization.
The overall framework of TraceLift-Groups and TraceLift . (a) Data curation pipeline of TraceLift-Groups . Then we use TraceLift-Groups to finetune the reward model specialized for reasoning supervisi…
Feed-forward network architecture of the ELAS. The input is first multiplied by the low-rank matrices of the up projection layer, then passes through the ReLU 2 \( \text{ReLU}^{2} \) activation function. The activation is applied with 2:4 structured sparsity and then multiplied with the low-rank matrix of the down layer using sparse matrix multiplication to obtain the output of the FFN layer.
Feed-forward network architecture of the ELAS. The input is first multiplied by the low-rank matrices of the up projection layer, then passes through the ReLU 2 \( \text{ReLU}^{2} \) activation functi…
cs.AIarxiv:2605.03667v1

ELAS: Efficient Pre-Training of Low-Rank Large Language Models via 2:4 Activation Sparsity

Jiaxi Li, Lu Yin et al.

ELAS proposes a novel framework for efficient large language model (LLM) pre-training by combining low-rank adaptation with 2:4 structured sparsity applied specifically to the activation matrices. This addresses the memory bottleneck caused by full-rank activa…

abstractpdfLINNET · SCORE9RELEVANCE
cs.AIarxiv:2605.03986v1

From Intent to Execution: Composing Agentic Workflows with Agent Recommendation

Kishan Athrey, Ramin Pishehvar et al.

This paper introduces an automated framework to compose Multi-Agent Systems (MAS) directly from a user's intent, replacing manual planning and agent selection. The core method involves an LLM-derived planner generating tasks, which are then mapped to suitable …

abstractpdfLINNET · SCORE9RELEVANCE
Architecture for an end-to-end MAS with dynamic and redundant workflow
Architecture for an end-to-end MAS with dynamic and redundant workflow
№06
cs.AI
9

MEMTIER: Tiered Memory Architecture and Retrieval Bottleneck Analysis for Long-Running Autonomous AI Agents

Bronislav Sidik, Lior Rokach

MEMTIER introduces a tripartite, tiered memory architecture to combat memory degradation in long-running AI agents, addressing failure modes in flat-file systems. Its core method i…

abstractpdf
№07
cs.AI
9

MOSAIC-Bench: Measuring Compositional Vulnerability Induction in Coding Agents

Jonathan Steinberg, Oren Gal

MOSAIC-Bench addresses the vulnerability of coding agents that comply with sequenced, innocuous requests to produce exploitable code, a weakness missed by isolated safety evaluatio…

abstractpdf
№08
cs.AI
9

OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories

Yuwen Du, Rui Ye et al.

OpenSeeker-v2 demonstrates that a simple Supervised Fine-Tuning (SFT) approach can effectively train powerful search agents, challenging the need for resource-intensive pipelines l…

abstractpdf
№09
cs.AI
9

OracleProto: A Reproducible Framework for Benchmarking LLM Native Forecasting via Knowledge Cutoff and Temporal Masking

Yiding Ma, Chengyun Ruan et al.

OracleProto introduces a reproducible framework to rigorously benchmark the native forecasting ability of Large Language Models (LLMs). It achieves this by reconstructing resolved …

abstractpdf
№10
cs.AI
9

QKVShare: Quantized KV-Cache Handoff for Multi-Agent On-Device LLMs

Pratik Honavar, Tejpratap GVSL

QKVShare introduces a framework for efficient, quantized Key-Value (KV) cache handoff between agents in on-device multi-agent LLMs. It utilizes token-level mixed-precision allocati…

abstractpdf
See all 20 papers →
§ III

The Town Square

Hacker News 10
590
pts
Top story
Accelerating Gemma 4: faster inference with multi-token prediction drafters

Google is accelerating Gemma 4's inference speed by implementing multi-token prediction drafters, allowing the model to generate multiple tokens simultaneously.

blog.google5 May discuss on HN →
525
AI didn't delete your database, you did
idiallo.com5 May
461
Three Inverse Laws of AI
susam.net5 May
395
Zuckerberg 'Personally Authorized and Encouraged' Meta's Copyright Infringement
variety.com5 May
342
California farmers to destroy 420k peach trees following Del Monte bankruptcy
sfgate.com5 May
249
AI Product Graveyard
tooldirectory.ai5 May
160
Telus Uses AI to Alter Call-Agent Accents
letsdatascience.com6 May
151
GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents
arxiv.org5 May
149
Zuckerberg 'personally authorized' Meta's copyright infringement, publishers say
apnews.com5 May
122
Show HN: Airbyte Agents – context for agents across multiple data sources
5 May
compiled overnight by google/gemini-2.5-flash-lite-preview-09-2025 · end of issue no. 9 · thank you for reading