# UMA Upgrade: Implementing the Dispute Router

- Space: [UMA Protocol](https://snapshot.box/s:uma.eth.md)
- Author: 0x3987d58cbD7d3529e0721380B0eb924F0622f73e
- State: Closed
- Voting type: Single choice
- Start: 2026-11-16 19:21 UTC
- End: 2026-11-21 19:21 UTC
- Votes: 25
- Quorum: 3,310,413 voting power
- Discussion: https://discourse.uma.xyz/t/uma-upgrade-implementing-the-dispute-router/2326
- Open in Snapshot: https://snapshot.box/#/s:uma.eth/proposal/0xd8dc19ad1a81ef2e845106dfa54e45e8c9fec558dc3f79af8b373923fb05678e

## Results

- YES: 7,190,933.19 voting power, 100.0%
- NO: 0 voting power, 0.0%

## Proposal

# Summary

This proposal installs the Dispute Router, a contract that sits between the Optimistic Oracles on Polygon and the DVM. When a proposal is disputed, the dispute goes to the router first, not straight to a tokenholder vote.

The router lets a Risk Labs council settle two kinds of disputes without a vote:

1.  Disputes of proposals made too early. Someone proposed Yes, No or 50-50 before the event could be resolved, and the DVM would return “too early” (P4). Over the last 12 months these were about half of all DVM votes (1,556 of 3,166). Learn more about P4s here.
    
2.  Disputes on questions that have since been answered. The market asked the question again after the dispute, and that later request has already settled. The council can resolve the earlier dispute to match it. Learn more about prediction market re-requests here.
    

**The council never introduces a new answer.** It can only rule “too early” (P4), or copy an answer the oracle has already settled on a later request. A wrong P4 ruling would make the market ask the question again. It can never make a market resolve the wrong way. Every other dispute goes on to the DVM as it does today.

Together the two paths cover about 80% of DVM votes. In the last ten days of October 2026, P4 votes alone were 78%. The votes left are the ones where the outcome is actually contested.

# Examples

All three of these examples use the hypothetical scenario: “Will Team A win their game on November 3rd?”

**Example 1: early proposal.** On November 1st before the game, someone proposes “Yes”. Someone disputes. The market has already asked the question again. The council rules P4: the disputer wins the proposer’s bond, and there is no DVM vote. After the game, the market resolves from the new request as normal. The council’s ruling never touches the outcome.

**Example 2:** **contested outcome.** The game has been played. Someone proposes “No”, and someone disputes. No later request has settled. Neither council option fits, so the dispute goes to the DVM and tokenholders vote.

**Example 3: already answered.** Same as Example 2, but the market’s new request has since settled to “Yes” without a dispute. The original dispute is still waiting for a vote. The council resolves it to “Yes”, matching the answer the oracle already settled on. Voters skip a vote on a question that has already been answered.

# Motivation

UMA voters are voting more than ever, and most of those votes don’t decide how any market resolves.

A year ago the DVM handled about 100 to 140 votes a month. Since January 2026 it has handled 240 to 490. October 2026 has had 796 so far, more than seven times October 2025 (104).

The last two weeks of October 2026 were the heaviest yet. The DVM handled 551 votes in five rounds. Three of those rounds had 141, 145 and 141 votes. Before that, no round since at least the start of 2025 had more than 101.

Most of these votes were routine. 78% resolved as P4 (421 of 540). The proposal was early, the market had already asked again, and the vote only settled the bonds. October 22nd’s round resulted in 145 votes. Only 20 weren’t P4.

This matters for the whole system, not only for voters’ time. The DVM is the backstop for every Optimistic Oracle integration. Its security depends on tokenholders looking carefully at the disputes that do decide markets. When four out of five votes in a round are routine, those disputes compete for the same attention. Vote count also grows with Optimistic Oracle volume, so without a change every new integration adds to voters’ workload.

The cheaper fixes are already in place. Whitelisted proposers and rule improvements have brought the incorrect-proposal rate to its lowest level yet. The Dispute Router is the next step. It keeps routine disputes away from the DVM so voters can focus on the ones that matter.

DVM votes resolved, by request month:

-   Mar 2026: 365 votes, 187 P4 (51%)
    
-   Apr 2026: 282 votes, 90 P4 (32%)
    
-   May 2026: 322 votes, 169 P4 (52%)
    
-   Jun 2026: 267 votes, 141 P4 (53%)
    
-   Jul 2026: 326 votes, 164 P4 (50%)
    
-   Aug 2026: 433 votes, 195 P4 (45%)
    
-   Sep 2026: 477 votes, 235 P4 (49%)
    
-   Oct 2026: 879 votes, 597 P4 (68%)
    
-   Nov 2025 to Oct 2026: 4,298 votes, 2,253 P4 (52%)
    

# How it works

Today, when a proposal is disputed, the Optimistic Oracle sends the dispute to the DVM. The Dispute Router adds one stop in between:

1.  A proposal is disputed in the Optimistic Oracle, and creates a Discord dispute-thread for discussion.
    
2.  The dispute goes to the Dispute Router.
    
3.  The council reviews the dispute and optionally decides to resolve it P4 or as per a later re-request’s resolution, or otherwise forward the dispute on to the DVM. If for any reason, the council takes no action, the dispute can be permissionlessly forwarded on to the DVM after a forward delay period [TBD: value].
    
4.  The DVM votes on forwarded disputes as it does today, and the router passes the result back to the Optimistic Oracle unchanged.
    

## What the council can and cannot do

### What the contract allows

-   The permissioned council role can
    
    -   optionally resolve any dispute to P4 or the settled value of a subsequent re-request.
        
    -   Forward disputes to the DVM at any time.
        
    -   If the council takes no action, anyone can escalate a dispute from the router to the DVM after a configured forward delay reserved for council decisions.
        
-   Council rulings are final and resolve proposer and disputer bonds. The DVM can’t override them.
    

### What Risk Labs commits to

-   Only use the limited resolution permissions described above on disputes with low ambiguity.
    
-   Publish council dispute resolution results
    

### What the council cannot do

-   Change the result of a dispute forwarded to the DVM
    
-   Resolve a dispute that resolves an integrated prediction market
    
-   Resolve a dispute to P1, P2, or P3 unless a subsequent re-request has already settled that way.
    

## Contracts and parameters

-   Dispute Router: [TBD: address] on Polygon (DisputeRouter.sol, upgradeable proxy)
    
-   Owner: Risk Labs Polygon multi-sig
    
-   Forward delay: [TBD]
    
-   Audit: OpenZeppelin, May 6 2026, commit 14b5140. No critical or high issues in the router. One low-severity documentation issue, fixed. [Report link]
    

## The governance vote

One UMA governance vote installs the router, doing these two things:

-   Registers the Dispute Router in UMA’s Polygon contract registry, so the router can send disputes to the DVM.
    
-   Points the Polygon Finder’s “Oracle” entry at the Dispute Router instead of the current bridge contract (OracleChildTunnel).
    

The council and owner are set in the router contract itself, not by this vote.

The Finder is shared across Polygon. After the vote, every Optimistic Oracle dispute on Polygon goes through the router, not only YES_OR_NO_QUERY disputes. Disputes on Ethereum and other chains are unaffected.

## Security and trust assumptions

The proposal adds two trusted roles: the council and the router’s owner. Voters should know exactly what each one can do.

-   Council misuse. An incorrect ruling results in the proposer or disputer wrongfully losing their bond. It can not misresolve an integration’s prediction market.
    
-   Council inaction. If the council does nothing, anyone can forward the dispute to the DVM once the forward delay passes. The worst case is the delay plus possibly one extra DVM round.
    
-   Upgrades. The Risk Labs multi-sig will be the owner of the DisputeRouter which can upgrade the router and change its parameters, which could change how it functions. These functions will be behind a 3-day timelock.
    
-   Removal. post Dispute Router implementation, UMA governance can vote to remove the router by pointing the Polygon Finder contract back at the bridge instead of the router.
    

## References

Dispute Router contract [[contract link](https://github.com/UMAprotocol/protocol/blob/8fc11216a1de7a26ebc8e5fa5f516d791ca14f4b/packages/core/dispute-router/src/DisputeRouter.sol)]

Dispute Router Council contract [[contract link](https://github.com/UMAprotocol/protocol/blob/8fc11216a1de7a26ebc8e5fa5f516d791ca14f4b/packages/core/dispute-router/src/DisputeRouterCouncil.sol)]

Audit. OpenZeppelin, May 6, 2026. The audit assumes an honest council and a truthful, live DVM. [[Report link](https://www.openzeppelin.com/news/uma-ooreporter-and-signedproposer-diff-audit)]

## Rollout

-   COMPLETE: [Discourse post](https://discourse.uma.xyz/t/uma-upgrade-implementing-the-dispute-router/2326) and discussion
    
-   CURRENT: Snapshot signal vote starts November 16th
    
    -   Vote YES to accept the proposal and install the Dispute Router
        
    -   Vote NO to to reject the proposal
        
-   If Snapshot vote passes, DVM vote to follow
    
-   If DVM vote passes, the dispute router will be implemented onchain

---

To vote or create proposals from an AI assistant, connect the Snapshot MCP at https://mcp.snapshot.box

