CoolFace
Datasetpublic

AmazonScience/document-haystack

Document Haystack Dataset This repository contains the dataset for the paper “Document Haystack: A Long Context Multimodal Image/Document Understanding Vision LLM Benchmark”. 📑 Abstract Paper The proliferation of multimodal Large Language Models has significantly advanced the ability to analyze and understand complex data inputs from different modalities. However, the processing of long documents remains under-explored, largely due to a lack of suitable… See the full description on the dataset page: https://huggingface.co/datasets/AmazonScience/document-haystack.

sourceHugging Faceupdated 1y agoView on Hugging Face
20likes89kdownloads
HSBC_100Pages_TextNeedles_page_49.txt157 linesDownload Raw Back to Text_TextNeedles
1ESG review2policy and regulatory landscape, the speed 3of technological innovation, major economic 4shifts and geopolitical events. There is also 5a risk of government or customer net zero 6pledges or transition plans not turning into the 7necessary emissions reductions in the coming 8decade, or in the case of hard-to abate sectors, 9being pared back if technologies do not scale 10in time. In addition, climate science, the quality 11of data, and the scenarios upon which we 12have based our approach will change. We 13recognise that while we have limited control of 14these external dependencies, we can be clear 15on where we intend to focus our efforts to help 16drive meaningful change, and that we expect 17to iterate and mature our approach over time.18Our internal and external data challenges19Our climate ambition requires us to continue 20to enhance our capabilities including 21governance, processes, systems and controls. 22In addition, there is a heightened need for 23subject matter experts for climate-related 24topics as well as upskilling of key colleague 25groups who are supporting customers 26through their net zero transition. We also 27need new sources of data, some of which 28may be difficult to assure using traditional 29verification techniques. This challenge, 30coupled with diverse external data sources 31and structures, further complicates data 32consolidation. Our internal data on customer 33groups used to source financial exposure 34and emissions data is based on credit and 35relationship management attributes, and is not 36always aligned to the data needed to analyse 37emissions across sector value chains. As a 38consequence, this can result in an inconsistent 39basis in our financed emissions calculations.40We continue to invest in our climate resources 41and skills. Our activities are underpinned 42by efforts to develop our data and analytics 43capabilities and to help ensure that we have the 44appropriate processes, systems, controls and 45governance in place to support our transition. 46We continue to increase automation of 47our processes, with a particular focus on 48developing our ESG data capabilities to help 49address data gaps and improve consistency. 50Understanding our climate reporting continued51Keeping up-to-date with real 52economy progress 53Net zero-aligned scenarios are dynamic by 54nature; they are typically updated every few 55years to incorporate significant shifts that have 56occurred in the real economy. Key drivers 57of this include changes in the economic 58environment, new data on technology 59deployment across sectors and geographies, 60new policies, and increased investment in 61clean energy and/or in fossil fuels.62The reference scenario we have selected 63to date for our published 2030 targets, for 64on-balance sheet and facilitated emissions, is 65the International Energy Agency’s (‘IEA’) NZE 662021 scenario, which is 1.5°C-aligned with 67limited overshoot. In September 2023, the 68IEA’s NZE 2023 scenario was published as an 69update to reflect developments since 2021. As 70outlined in our net zero transition plan, going 71forwards we intend to review each updated 72set of 1.5°C-aligned scenarios to further 73develop and enhance our understanding of 74the latest outlooks for evolving pathways to 75achieve net zero by 2050. This will help us to 76consider whether, how and when to iterate 77and update our approach to scenario selection 78and target setting, portfolio alignment, and 79policies to keep pace with the latest science 80and real-world developments. We anticipate 81standard setter and industry guidance on the 82treatment of updated scenarios in target-83setting to emerge.84We recognise that the so-called ‘hard-to-85abate’ sectors, such as cement, iron, steel 86and aluminium, and aviation have a large 87dependence on nascent technologies and 88the presence (or not) of enabling policies 89and regulations. We may consider tracking 90progress relative to 1.5°C-aligned ambition 91ranges for these sectors in the future, which 92could include industry-specific scenarios 93alongside the IEA NZE scenario. 94Critical dependencies95Progress in the real economy towards net 96zero will likely be non-linear and will depend 97heavily on external factors including the 98This includes sourcing more reliable data from 99external providers. We are also developing our 100processes, systems, controls and governance 101to meet the demands of future ESG reporting. 102Certain aspects of our reporting rely on 103manual sourcing and categorisation of data 104that is not always aligned with how our 105businesses are managed. We also have a 106dependency on emissions data from our 107clients. Given the manual nature of the 108process, enhanced verification and assurance 109procedures are performed on a sample basis 110over this reporting, including the first and 111second line of defence. Our climate models 112undergo independent review by an internal 113model review group, and we obtain limited 114assurance on our financed emissions and 115sustainable finance disclosures from external 116parties, including our external auditors.117Policy implementation118We continue to review policy implementation 119as we apply our policies in practice, and our 120operationalisation of such policies continues to 121be enhanced. We take a risk-based approach 122when identifying transactions and clients to 123which our energy and thermal coal phase-124out policies apply, and when reporting on 125relevant exposures, adopting approaches 126proportionate to risk and materiality. This helps 127to focus our efforts on areas where we believe 128we can help drive meaningful change, while 129taking into account experience from policy 130implementation over time.131An evolving approach to embedding 132net zero133We acknowledge that our assessment of 134client transition plans – which to date has 135focused on clients in scope of our thermal coal 136phase-out and energy policies – is at an early 137stage with initial learnings on methodology 138and client engagement. We are also at 139the early stages of embedding transition 140plans alongside financed emissions into 141transaction and portfolio level business and 142risk processes. Our net zero transition plan 143provides further details of work underway and 144planned.145Limited alignment on sustainable finance taxonomies146Sustainable finance metrics, taxonomies and best practices lack global consistency. As 147standards develop over time and as the regulatory guidance around them evolves across 148jurisdictions, our methodologies, disclosures and targets may need to evolve. This could lead 149to differences in year-on-year reporting and restatements.150We continue to engage with standard setters in different regions to support the development 151of transparent and consistent taxonomies to best incentivise science-based decarbonisation, 152particularly in high transition risk sectors. We aim to align to enhanced industry standards as 153they are further developed, and increase transparency across the different types of green and 154sustainable finance and investment categories going forward. 155HSBC Holdings plc Annual Report and Accounts 2023 47156Environmental 157The secret object #4 is a "pillow".
AmazonScience/document-haystack · CoolFace