We present Subject Fidelity Optimization (SFO), a novel comparative learning framework for zero-shot subject-driven generation that enhances subject fidelity. Existing supervised fine-tuning methods, which rely only on positive targets and use the diffusion loss as in the pre-training stage, often fail to capture fine-grained subject details. To address this, SFO introduces additional synthetic negative targets and explicitly guides the model to favor positives over negatives through pairwise comparison. For negative targets, we propose Condition-Degradation Negative Sampling (CDNS), which automatically produces synthetic negatives tailored for subject-driven generation by introducing controlled degradations that emphasize subject fidelity and text alignment without expensive human annotations. Moreover, we reweight the diffusion timesteps to focus fine-tuning on intermediate steps where subject details emerge. Extensive experiments demonstrate that SFO with CDNS significantly outperforms recent strong baselines in terms of both subject fidelity and text alignment on a subject-driven generation benchmark. Project page: https://subjectfidelityoptimization.github.io/

Keyphrase ranking plays a crucial role in information retrieval and summarization by indexing and retrieving relevant information efficiently. Advances in natural language processing, especially large language models (LLMs), have improved keyphrase extraction and ranking. However, traditional methods often overlook diversity, resulting in redundant keyphrases. We propose a novel approach using Submodular Function Optimization (SFO) to balance relevance and diversity in keyphrase ranking. By framing the task as submodular maximization, our method selects diverse and representative keyphrases. Experiments on benchmark datasets show that our approach outperforms existing methods in both relevance and diversity metrics, achieving SOTA performance in execution time. Our code is available online.

A word or phrase that provides information necessary to fulfill a particular intent. Each entity belongs to a category specified by the entity's associated type. For instance, a book_flight intent could have a location type for entities like'Miami' and'Chicago O'Hare', an airline type for entities like'Air India' and'Southwest', and a date type for entities like'July 4th' and'New Years Day'. An application-agnostic entity that is automatically detected by MindMeld. Examples include numbers, time expressions, email addresses, URLs and measured quantities like distance, volume, currency and temperature.

A new generation of automated bin picking systems using deep learning is evolving to support increasing demand for e-commerce. To accommodate a wide variety of products, many automated systems include multiple gripper types and/or tool changers. However, for some objects, sequential grasp failures are common: when a computed grasp fails to lift and remove the object, the bin is often left unchanged; as the sensor input is consistent, the system retries the same grasp over and over, resulting in a significant reduction in mean successful picks per hour (MPPH). Based on an empirical study of sequential failures, we characterize a class of "sequential failure objects" (SFOs) -- objects prone to sequential failures based on a novel taxonomy. We then propose three non-Markov picking policies that incorporate memory of past failures to modify subsequent actions. Simulation experiments on SFO models and the EGAD dataset suggest that the non-Markov policies significantly outperform the Markov policy in terms of the sequential failure rate and MPPH. In physical experiments on 50 heaps of 12 SFOs the most effective Non-Markov policy increased MPPH over the Dex-Net Markov policy by 107%.

The Serious Fraud Office is a specialist prosecuting authority tackling the top level of serious or complex fraud, bribery, and corruption across England, Wales and Northern Ireland. It both investigates and prosecutes its cases, which is unique but necessary because the cases are complicated, and lawyers and investigators need to work together from the outset of them. Supporting both lawyers and investigators is an IT infrastructure that is managed by chief technology officer (CTO) Ben Denison, who also looks after the technology that supports the SFO's operational activities. "That includes digital evidence, so in our cases when we're investigating someone we'll see information from phones, tablets, laptops, and email, and we have a digital forensics team that processes that material and extracts relevant data from it, and it can then be ingested into another system which our case teams use to review it from," Denison tells PublicTechnology. In essence, the SFO ensures that all of the data – whether it's digital or a hard copy – is put into one place, so that it can be reviewed together.

It previously piloted similar technology developed by Canadian firm OpenText during its four-year investigation into fraud at Rolls-Royce which involved reviewing 30 million documents. The SFO said that technology was up to 80% cheaper than using outside counsel to review those documents and identify legally privileged material. OpenText, the "AI lawyer", goes "further than just flagging legally privileged material" an SFO spokesperson told Sky News. "It can also scan and organise information from multiple document types - PowerPoint, Outlook calendar invites, Word documents etc - displaying the information relevant to an investigation on a timeline for an investigator to then review." The SFO told Sky News they expect the system to cost "around £12m over the expected lifetime of 7 years - which is offset against the savings the new tech will bring by enhancing our ability to review and investigate in a targeted way, without solely relying on human review."

Robot investigators could be widely used in future to examine documents in complex cases, the head of the Serious Fraud Office (SFO) has suggested. David Green said he would like to see the possibility of employing artificial intelligence "carefully examined" after using technology to sift through a cache of 30 million documents disclosed by Rolls-Royce during a major investigation. He also said it was now "pretty clear" that his agency would continue as an independent body after the government dropped plans to have it taken into the National Crime Agency. The SFO director set out how the Rolls-Royce documents had been examined by a computer algorithm which had the ability to learn as it went along. The technology was trying to find legally privileged documents which could not be used in the case, but Mr Green suggested that in future similar methods could be used to identify useful evidence in investigations.

The Serious Fraud Office (SFO) had a problem. Its investigation into corruption at Rolls-Royce was inching towards a conclusion, but four years of digging had produced 30 million documents. These needed to be sorted into "privileged" and "non-privileged", a legal requirement that involves paying junior barristers to do months of repetitive paperwork. "We needed a way that was faster," says Ben Denison, chief technology officer at the SFO. So, in January 2016, he started working with RAVN.