Bounded news overlays for risk-aware asset allocation
Don’t ask one language model for the trade. Ask small models narrow questions, and let a risk model allocate.
A common way to build a news-reading trading agent is to hand the news to a large language model and let it produce the trade, an overlay or an event signal directly. AlphaAllocator splits that job into the two questions news can answer: which asset is this about, and is it good or bad, each with its own small model. A classical risk model decides the allocation, and the news may only nudge it within a fixed budget.
Details
Please help me understand AlphaAllocator, a research agent built for NLPCC 2026 Shared Task 4.
Project page: https://qrsntu.org/projects/alphaallocator/
How it works: https://qrsntu.org/projects/alphaallocator/details/
Plain-text summary: https://qrsntu.org/projects/alphaallocator/summary.md
Paper: https://openreview.net/forum?id=u09PzTYNHk
Design
AlphaAllocator does not hand the news to one large language model and ask it for the trade. It splits the job into the two questions news can answer, each with its own small model: a relevance model (a 4-layer BGE-family sentence encoder) decides which asset a news item is about, and a tone model (FinBERT) decides whether it is good or bad news. Their product, weighted for recency and certainty and centred across assets, gives one news score per asset. A risk-parity model decides the allocation (in Track 2, a blend of inverse-volatility, momentum and trend), and the news may only tilt it: the whole tilt is capped at 1.0% of the portfolio (1.2% in Track 2), so prices decide and news only nudges.
Key results
- NLPCC 2026 Shared Task 4 (LLM-based Investment Advisor Agents for Asset Allocation in the Chinese Market), scored by the organizers on a hidden 2026 period.
- Track 1 (11 macro assets): 1st of 10 teams. Sharpe 1.1144, annualised return 13.52%, max drawdown −6.23%.
- Track 2 (16 sector assets): 6th of 13 teams. Sharpe 0.7566, annualised return 15.32%, max drawdown −7.49%.
- Public-2025 replay of Track 1 (243 daily decisions): Sharpe 2.270 with news against 2.198 for the same agent with the news zeroed (return 24.30% against 23.31%).
- ACM ICAIF 2026 Trading Agent Competition (30 U.S.-listed equities): results coming soon.
One day’s news, step by step: link each item to the assets it is about, mark it good or bad, pool it into one score per asset, then let the risk model allocate while the news nudges within a capped budget.Each news item published before the cutoff is linked to the assets it is about and marked good or bad; the pooled scores nudge a risk-model portfolio within a capped budget.Illustrative example; the agent reads up to 20 items a day.
AlphaAllocator at a glance
The idea
For news to help a portfolio, an agent has to know two things: which asset an item is about, and whether it is good or bad news. We give each question its own small model and multiply the answers:
The scores, weighted for recency and certainty, give one score per asset. They tilt a risk-parity portfolio by at most = 1.0% of it in total:
Every allocation the news can reach. Centred, and scaled down whenever it would exceed the cap, so three assets let the whole reachable region be drawn in the plane: a hexagon, each corner moving 0.5% out of one asset and 0.5% into another. The inset sets it against every portfolio of the three, at true scale: all of it is smaller than the dot marking today’s portfolio. Clipping and rescaling afterwards cannot enlarge it. Drag the news, turn it up, or raise every score equally.Illustrative assets and scores: the geometry the equation fixes, not a measurement.
Experiments
NLPCC 2026 · Chinese funds tested
Track 1
Sharpe
Ann. return
Max DD
1st
1.1144
13.52%
−6.23%
2025 replay Sharpe, with news2.270same agent, news zeroed2.198
Track 1: 11 macro assets, 10 teams, organizer-run on a hidden 2026 period (next-best Sharpe 0.7792). Track 2 (sectors): 6th of 13.
ICAIF 2026 · US equities Coming soon
The same design, adapted to 30 US stocks and 7 decisions a day, is in the ACM ICAIF 2026 Trading Agent Competition.
Results coming soon.
Every team’s official Sharpe, NLPCC 2026. The organizers’ hidden 2026 evaluation, in rank order, ours in teal. Track 1: 1st of 10. Track 2: 6th of 13. More results →