Get Started
Topic · 0 recaps
Business
Content about business strategies, corporate news, management practices, entrepreneurship, and industry trends.
...
Posts
Questions
Home
Topics
Search
Library
Questions researchers are working on
Follow a question through Rcap’s explanations and the latest papers addressing it.
Search
How can advertising enter token-by-token generated responses while preserving incentive compatibility and response quality?
Generated responses do not offer the fixed advertising slots assumed by conventional mechanisms, so advertising may need to influence the generation process itself. That influence must not compromise truthful advertiser participation or the usefulness of the response.
How can B2B marketing teams resolve fragmented company records to predict conversion over long sales cycles?
B2B buying processes can span months or years, while one company may appear across fragmented contact records and inconsistent company names. This makes it difficult to construct a reliable customer-level view and determine which prospects are likely to convert.
How can candidate generation recover corporate-family links between supplier records with no shared name evidence?
Suppliers in the same corporate family may be deliberately represented as different entities, while the relationship may appear in neither record. When their names share no distinctive evidence, blocking can exclude the true pair before matching is applied.
How can climate disclosure classifiers remain reliable when documents shift across sources?
Annual reports, press releases, and earnings calls differ in length, purpose, and writing style. Adaptation that performs well on one source may therefore lose effectiveness when applied to another.
How can corporate distress be forecast from high-dimensional, mixed-frequency data with right-censored outcomes?
Distress data may show that a firm has not failed by the study’s end without revealing its eventual event time. Numerous predictors sampled at different frequencies further complicate estimation and uncertainty quantification.
How can e-commerce recommenders distinguish genuinely complementary products from items merely bought together?
Purchase logs capture correlation, not whether products function together, so frequent co-purchases can produce irrelevant basket-building suggestions. The difficulty is inferring functional compatibility from noisy, large-scale behavioral data.
How can e-commerce search combine text, images, and voice to understand product and purchase intent?
Separate text, visual, and voice search systems cannot consistently interpret queries that span modalities. Generic multimodal models may also miss the domain knowledge and purchase reasoning needed for precise product retrieval.
How can enterprises determine whether an AI agent meets reliability targets at acceptable oversight and operating cost?
Benchmark task-completion scores do not show whether an agent can satisfy a workflow’s reliability target in practice. Deployment decisions also depend on the human review required and the cost of operating the human–AI system.
How can enterprises turn tacit, fragmented knowledge into grounded, adaptable, auditable actions?
Organizational know-how is distributed across structured records, documents, multimodal material, and tacit practices, while generic models lack firm-specific decision context. Retrieval systems can remain brittle, and static playbooks cannot adapt or directly support governed execution.
How can governed language-model analytics preserve expressive queries while producing replayable, evidence-backed answers?
Enterprise users want natural-language access to analytical operations such as aggregation, comparison, windows, ranking, and similarity. Runtime planning can make execution and supporting evidence inconsistent, while overly restrictive governance may narrow the supported query class.
How can human reviewers reliably detect LLM errors when verification reasoning is hard to retrieve?
Reviewers may miss LLM errors even when they understand how to verify outputs, because the relevant reasoning is not always accessible when review occurs. Repeated LLM use can make this difficulty more consequential.
How can joint replenishment be coordinated in real time across many heterogeneous items?
Joint ordering couples thousands of item decisions through demand uncertainty, lead times, and shared costs. Optimization can become slow at scale, while learning systems must assign credit across a high-dimensional action space.
How can LLMs adapt coupled retail supply-chain pipelines when changing requirements have multiple routes with different downstream effects?
A new operational requirement may be satisfied by changing different modules in a coupled decision pipeline, with each route affecting downstream outcomes differently. The adaptation must preserve admissibility and end-to-end operational performance.
How can models assign credit to individual steps from trajectory feedback?
Feedback on a completed task often leaves unclear which steps helped. Some systems have verified outcomes to learn from; others must work from rubric judgments that estimate quality.
How can organizations adopt generative AI when technical reliability, social risks, and governance lag behind?
Organizations must weigh generative AI’s potential to improve productivity and innovation against unreliable outputs, ethical risks, and incomplete governance. These difficulties reflect a mismatch between rapidly changing technical capabilities and slower-adapting organizational and societal systems.
How can return narratives guide inspection depth and recovery allocation under uncertain asset condition and limited labor?
Returned assets must be routed or recovered before their true condition is known, while inspecting every item can exceed available labor. Narrative notes may contain useful but unevenly reliable clues, making inspection depth and recovery allocation difficult to determine.
How can we forecast evolving product-attribute preferences from users’ time-ordered interactions?
Users may repeatedly reveal preferences for attributes such as brand, size, or color, but those preferences can evolve over time. The challenge is to infer future attribute choices from this temporal history at a fine-grained level.
How can we identify which inputs genuinely drive an opaque sales forecast?
Accurate deep sales forecasters can conceal which signals drive individual predictions. Post-hoc attribution methods may produce plausible explanations that do not reflect the model’s actual behavior.
How should enterprise decision agents be evaluated when rankings change between fixed-opponent and shared-market settings?
An agent may perform well against fixed opponents yet behave differently when other evaluated agents compete for shared resources. Evaluation design must therefore separate apparent model quality from effects caused by the surrounding competitive ecology.
How should forecasting-model selection adapt to demand patterns, available history, and forecast horizon?
A single model-selection rule may choose inconsistently across demand patterns and forecasting conditions. The suitable selector can change with both the amount of historical data and how far ahead forecasts are required.
Previous
1 / 2
Next