LLM (Large Language Model)
Search documents
Cerence(CRNC) - 2026 Q1 - Earnings Call Transcript
2026-02-04 22:30
Cerence (NasdaqGS:CRNC) Q1 2026 Earnings call February 04, 2026 04:30 PM ET Speaker4Good day, and thank you for standing by. Welcome to the Cerence First Quarter 2026 earnings call. At this time, all participants are in listen-only mode. After the speaker's presentation, we'll open up for questions. To ask a question during the session, you will need to press star 11 on your telephone. You will then hear an automated message advising your hand is raised. To withdraw your question, please press star 11 again ...
野村 2025 年投资论坛:AI 相关主题 -与 Sakana AI 及 Ibiden 的对话-Nomura Investment Forum 2025_ AI-related themes_ Conversation with Sakana AI and Ibiden
野村· 2025-12-08 00:41
Investment Rating - The report assigns a "Buy" rating to Ibiden, indicating a positive outlook for the company's performance in the market [1]. Core Insights - The AI industry is expected to see sustainable growth through application-specific custom AI and packaging technologies that lower inference costs [1]. - Sakana AI focuses on developing innovative algorithms and application-specific AI models, aiming for sustainable growth by specializing in niche areas [2]. - Ibiden is increasing its market share in the custom ASIC packaging sector, with manufacturing difficulties for ASICs approaching those of GPUs, which may lead to a broader customer base for AI accelerator packages [3][5]. Summary by Sections Sakana AI - Sakana AI, founded by notable figures from Google Brain and Mercari Europe, specializes in advanced algorithms and application-specific AI models [2]. - The company aims to improve algorithms efficiently through an evolutionary process rather than relying on large-scale computing resources [2]. - Concerns about sustainability in ultra-large model learning persist, as only a few companies can maintain the necessary infrastructure [2]. Ibiden - Ibiden has refined its manufacturing technologies for large packages since the mid-2010s, with expectations of increased manufacturing difficulty for custom ASICs and GPUs [3]. - The company is likely to expand its customer base for AI accelerator packages due to its capability to stably mass produce EMIB-T, a technology that may see wider adoption by 2027-28 [3]. - Technological advancements in providing LLM inference with stable quality at lower prices are expected to enhance Ibiden's competitive advantage in the market [3][5].
X @Avi Chawla
Avi Chawla· 2025-10-12 19:29
Core Problem of Traditional RAG - Most retrieved chunks in traditional RAG setups do not effectively aid the LLM, leading to increased computational costs, latency, and context processing [1][5] - Classic RAG involves fetching similar chunks from a vector database and directly inputting the retrieved context into the LLM [5] REFRAG Solution by Meta AI - Meta AI's REFRAG introduces a novel approach by compressing and filtering context at a vector level, focusing on relevance [1][2] - REFRAG employs chunk compression, relevance policy (RL-trained), and selective expansion to process only essential information [2] - The process involves encoding documents, finding relevant chunks, using a relevance policy to select chunks, and concatenating token-level representations [3][4] Performance Metrics of REFRAG - REFRAG outperforms LLaMA on 16 RAG benchmarks, demonstrating enhanced performance [5][7] - REFRAG achieves 30.85x faster time-to-first-token, significantly improving processing speed [5][7] - REFRAG handles 16x larger context windows, allowing for more extensive information processing [5][7] - REFRAG utilizes 2-4x fewer tokens, reducing computational resource consumption [5][7] - REFRAG leads to no accuracy loss across RAG, summarization, and multi-turn conversation tasks [7]
Taboola.com (TBLA) FY Conference Transcript
2025-08-12 18:15
Summary of Taboola.com (TBLA) FY Conference Call - August 12, 2025 Company Overview - **Company**: Taboola.com (TBLA) - **Industry**: Performance Advertising in the Open Web - **Market Opportunity**: $55 billion market opportunity in performance advertising [4][5] Core Business Model - **Unique Offering**: Taboola is a leading performance advertising platform that complements search and social advertising by providing targeted ads based on first-party data [3][4] - **Daily Reach**: The company reaches approximately 600 million people daily through partnerships with major publishers like Yahoo, Apple News, Disney, and NBC [4] - **Revenue Goals**: Targeting $2 billion in revenue from a $55.7 billion market, with over $200 million in adjusted EBITDA, representing a margin of over 30% [5] Financial Performance - **EBITDA Margin**: The company maintains a strong EBITDA margin of over 30% and a free cash flow of 70% of EBITDA, which is being used for share buybacks [5][67] - **Share Buybacks**: Taboola has repurchased 12% of its shares in the first half of the year and plans to continue aggressive buybacks [5][69] Market Position and Strategy - **Two-Sided Marketplace**: Taboola operates a two-sided marketplace with exclusive long-term relationships with 11,000 publishers, providing predictable inventory and access to consumer data [6][7] - **Shift to Performance Marketing**: The introduction of the Realize product marks a pivot towards broader performance marketing, allowing advertisers to use various ad formats beyond native advertising [12][14] - **Display Advertising Market**: Taboola estimates a $10 billion display ad market among its publishers, aiming to capture 30% market share [18] Growth and Future Outlook - **Growth Strategy**: The company aims to double its revenue from $2 billion to $4 billion primarily through increased demand and spending from advertisers [15][25] - **Realize Product Adoption**: Early signs of success with Realize include 650 advertisers trying the product, with existing advertisers increasing their spending [27][28] - **Focus on Performance Advertising**: Taboola is committed to performance advertising, avoiding branding-focused areas like CTV, which is seen as a competitive and less favorable market [36][39] Challenges and Market Dynamics - **Native Advertising Growth**: The native advertising space is not growing as expected, prompting the shift to a broader performance advertising strategy [22][23] - **Impact of Search Traffic**: Currently, only 5% of Taboola's traffic is driven by search, and the company has not seen significant impacts from changes in search dynamics [48][49] Technology and Innovation - **Use of AI and LLMs**: Taboola is leveraging machine learning and large language models (LLMs) across various departments to enhance productivity and create value [65][66] - **Predictive Audiences**: The company is developing features like predictive audiences to help advertisers optimize their campaigns [64] Conclusion - **Investment Philosophy**: Taboola prioritizes growth while maintaining profitability, with a focus on responsible business practices and maximizing shareholder value through buybacks and strategic investments [56][68] - **Future Expectations**: The company is optimistic about returning to double-digit growth through the successful adoption of Realize and continued investment in technology and partnerships [25][60]