FutureSim is a benchmark that replays real news from January to March 2026 for AI agents to forecast events, with top accuracy at 25% and some agents worse than no-prediction baselines on Brier skill score.
(1) Stage 1: correct analysis of the constituent parts of the problem question (a) Identify the FACTS given—place on a tree diagram. (b) Identify the primary and secondary LEGAL ISSUES raised by the facts, available defences and doubts in the law. Place on a tree diagram of the issues. List the issues under the facts. (c) Consider the LAW THAT MAY APPLY (eg, legislation or common law and/ or European Community law). The sources of law to be drawn on will vary according to the particular subject. Quickly list these under the issues on the tree diagram. (2) Stage 2: Begin to work on discrete aspects of the problem question (a) Decide the order in which issues will be raised in your answer. (b) Consider your view of uncertainties and gaps in the law in the area. (c) Consider issues of interpretation and defence. A doubt about the interpretation of the law is not a defence, it is a doubt about the law. Make sure you do not make this mistake, as they require a different approach. (3) Stage 3: Decide your view of the outcome of the specific questions asked in the problem question The facts in a problem question can give rise to many issues but all of these may not be necessary to resolve the specific question(s) set in your problem. Problem questions tend to ask you to do two main things: (a) Discuss the issues raised in the problem scenario. OR (b) Advise one of the parties. Both types of problem question require the same knowledge to successfully answer them. However, your approach will be different. (In fact essay questions can be drawn from the same knowledge but also require a different approach.) • In those drafted in response to a question in the style of (a) you raise all issues without privileging one party. • In those drafted in response to a type (b) question you raise all issues but orientate to your argument to the effect of those issues on the party you are asked to advise. This includes discussing in detail the likely chances of the other party being the successful party. 8.5.3 Demonstration: beginning to answer a specific problem question The key to successfully answering a problem question lies in spotting the ‘clues’ to the issues to be discussed. Many of these are purely linguistic. We will look at one particular problem, Problem Question 4, above, applying the stages outlined above.
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
FutureSim: Replaying World Events to Evaluate Adaptive Agents
FutureSim is a benchmark that replays real news from January to March 2026 for AI agents to forecast events, with top accuracy at 25% and some agents worse than no-prediction baselines on Brier skill score.