May 22, 2024

How aider scored SOTA 26.3% on SWE Bench Lite

Aider scored 26.3% on the SWE Bench Lite benchmark, achieving a state-of-the-art result. The previous top leaderboard entry was 20.3% from Amazon Q Developer Agent.

All of aider’s results reported here are pass@1 results, obtained without using the SWE Bench hints_text. All results in the above chart are unhinted pass@1 results. Please see the references for details on the data presented in this chart. It was corrected on 5/30/24 to reflect apples-to-apples comparisons, using pass@1 results from AutoCodeRover and results from OpenDevin that don’t use hints. The official SWE Bench Lite leaderboard only accepts pass@1 results that do not use hints.

Interactive, not agentic

Aider achieved this result mainly through its existing features that focus on static code analysis, reliable LLM code editing, and pragmatic UX for AI pair programming. Aider intentionally has quite limited and narrow “agentic behavior” to avoid long delays, high token costs and the need for users to repeatedly code review incorrect solutions. It’s also worth noting that aider currently does not use RAG, vector search, tools or give the LLM access to search the web or unilaterally execute code.

Aider is first and foremost an interactive tool for engineers to get real work done in real code bases using a chat interface. Aider provides a pair programming UX where users can ask for a change and see the edits performed in real-time. Aider can also offer additional help like fixing lint or test errors, but the user is always in full interactive control. This lets them quickly steer misunderstandings back on course and avoid wasting time and token costs.

Benchmark methodology

For the benchmark, aider was launched in each problem’s git repository with the problem statement submitted as the opening chat message from “the user.” After that aider runs as normal, with the following modifications:

Aider’s suggestions were always accepted without user approval.
A simple harness was used to retry the SWE Bench problem if aider produced code that wasn’t plausibly correct. Plausibly correct means that aider reported that it had successfully edited the repo without causing syntax errors or breaking any pre-existing tests.
If the solution isn’t plausible, the harness launches aider to try again from scratch, alternating between using aider with GPT-4o and Opus.
If no plausible solution is found after six tries, the harness picks the solution with the fewest edit/lint/test problems.

It’s important to be clear that aider and the benchmark harness only had access to the pre-existing tests in each problem’s repo. The held out “acceptance tests” were only used after benchmarking to compute statistics on which problems aider correctly resolved.

The full harness to run aider on SWE Bench Lite is available on GitHub.

The benchmarking process was similar to how a developer might use aider to resolve a GitHub issue:

They could launch aider in their repo with the command below, which tells aider they want to accept every suggestion and to use pytest to run tests.
- aider --yes --test-cmd pytest
They could start the chat by pasting in the URL or text of a GitHub issue. Aider will pull in the URL’s content and then try and solve the issue.
If aider doesn’t produce code that lints and tests clean, the user might decide to revert the changes and try again, maybe using aider with a different LLM this time. Aider is tightly integrated with git, so it’s always easy to revert AI changes that don’t pan out.

Outside a benchmark setting, it’s probably unwise or at least highly inefficient to let any AI agent run unsupervised on your code base. The reason aider is intended to be used interactively is so that the user can participate and direct aider’s work and approve suggestions. This way the user can offer immediate feedback or corrections if their initial instructions turn out to be ambiguous, or if the AI starts going down a wrong path.

Aider with GPT-4o alone was SOTA

Running the benchmark harness only using aider with GPT-4o to find plausible solutions achieved a score of 25.0%. This was itself matching the state-of-the-art, before being surpassed by the main result being reported here that used aider with both GPT-4o & Opus.

As noted below, a single attempt using Aider with GPT-4o tied the current top entry on the leaderboard.

Aider with GPT-4o & Opus

The benchmark harness alternated between running aider with GPT-4o and Opus. The harness proceeded in a fixed order, always starting with GPT-4o and then alternating with Opus until a plausible solution was found for each problem.

The table below breaks down the plausible solutions that were found for the 300 problems. It also provides details on the 79 that were ultimately verified as correctly resolving their issue. Some noteworthy observations:

Just the first attempt of Aider with GPT-4o resolved 20.3% of the problems, which ties the Amazon Q Developer Agent currently atop the official leaderboard.
Including the second attempt, Aider with GPT-4o and Opus scored 23.6% on the benchmark. These first two attempts obtained ~75% of all plausible and ~90% of all resolved solutions.
A long tail of solutions continued to be found using both models including one correctly resolved solution on the final, sixth attempt of that problem.

Attempt	Agent	Number of plausible solutions	Percent of plausible solutions	Number of correctly resolved solutions	Percent of correctly resolved solutions	Score on SWE Bench Lite
1	Aider with GPT-4o	208	69.3%	61	77.2%	20.3%
2	Aider with Opus	49	16.3%	10	12.7%	3.3%
3	Aider with GPT-4o	20	6.7%	3	3.8%	1.0%
4	Aider with Opus	9	3.0%	2	2.5%	0.7%
5	Aider with GPT-4o	11	3.7%	2	2.5%	0.7%
6	Aider with Opus	3	1.0%	1	1.3%	0.3%
Total		300	100%	79	100%	26.3%

If we break down the solutions solely by model, we can see that aider with GPT-4o outperforms Opus. This isn’t a fair and direct comparison, because GPT-4o always took the first turn and therefore got first crack at all the “easiest” problems. Aider with Opus only ever saw problems that GPT-4o failed to find plausible solutions for on its first try.

Aider with GPT-4o was producing higher quality plausible solutions, with a greater chance of going on to be accepted as resolving the issue. Again, this is biased by the turn ordering. But other anecdotal evidence from earlier runs of the benchmark also supports the observation that aider with GPT-4o is significantly stronger than Opus for this benchmark.

Agent	Number of plausible solutions	Number of correctly resolved solutions	Percent of plausible which correctly resolved
Aider with GPT-4o	239	66	27.6%
Aider with Opus	61	13	21.3%
Total	300	79	26.3%

Repository map, not RAG

The crucial first step in solving a SWE Bench problem is figuring out which parts of the repo are relevant and which files need to be edited. Most coding agents use some combination of RAG, vector search and providing the LLM with tools to interactively explore the code base.

Aider instead uses a repository map to help the LLM understand the layout, code structure, and content of a git repo. The repo map is created through static analysis of the code’s abstract syntax tree and call graph to provide a compact and powerful summary of the entire code base. The map is constantly tailored to show repo context that is relevant to the current state of the chat conversation. This is done by performing a graph optimization on the code’s call graph.

When the user asks for a change to their code, the LLM can use the repo map to decide which files to edit. The LLM simply returns a normal text response explaining which files it needs to edit and why. Aider notices when the LLM mentions filenames from the repo, and asks the user if they should be added to the chat. Adding a file to the chat allows the LLM to see the full contents of the file and edit it.

Please add a new /factorial/N endpoint.

To add a new /factorial/N endpoint, the most likely file that needs to be edited is app.py. Please add app.py to the chat so I can proceed with the changes.

app.py
Add these files to the chat? yes

This is a convenient and natural workflow for interactive chat, and it worked well for the SWE Bench problems. Aider successfully identified the correct file to edit in 70.3% of the benchmark tasks.

We can determine which file needs to be edited using the “gold” patch which is associated with each SWE Bench task. This patch was created by a human developer to solve the issue, and therefore reveals a file which can be edited to solve the problem. Of course aider is not able to see or use the gold patch or the file names it contains in any way. This information was only used to compute statistics outside the benchmarking process.

Reliable code editing

Once files have been selected for editing, the next step is of course to edit the source code to fix the problem.

Aider goes to great lengths to ensure that LLMs can not just write code, but reliably edit code. Aider has a collection of prompting strategies and code editing backends which have been honed through extensive benchmarking. These foundational capabilities help ensure that aider can properly integrate code from LLMs into an existing code base and source files.

The repository map helps here too, making sure that the LLM can see relevant classes, functions and variables from the entire repo. This helps ensure that the project’s existing APIs and conventions are respected and utilized when new code is added.

Regardless, there are still cases where aider may be unable to cleanly complete the edits specified by the LLM. This is usually because the LLM has failed to conform to the editing instructions in its system prompt. When aider completes, it returns an editing outcome that indicates whether it was able to successfully apply all edits. The benchmark harness uses this editing status as one criteria to determine if aider has created a plausible solution.

Linting and fixing

Another key criteria for a plausible solution is that it passes basic linting, which means that the code has no syntax or other fatal errors. Aider lints code after every LLM edit and offers to automatically fix any problems.

Aider ships with built-in linters based on tree-sitter which work with most popular programming languages. Aider shows linting errors to the LLM in a novel format, using the abstract syntax tree to display relevant code context for each error. This context helps LLMs understand the problem and make the correct changes to resolve it.

app.py:23:36: F821 undefined name 'num'  
  
app.py:  
...⋮...  
  6│class LongNum:  
...⋮...  
 19│    def expound(self, threshold):  
 20│        number = self.basis  
 21│        while number < threshold:  
 22│            number *= self.factor  
 23█        return num  
 24│  
 25│  
...⋮...  

Attempt to fix lint errors? yes

In the benchmark, these linting suggestions are always accepted. At completion, aider reports a linting outcome that indicates if it was able to produce code without any outstanding linting errors. The benchmark harness uses this status as one of the criteria to determine if aider has created a plausible solution.

Testing and fixing

The final crtieria for a plausible solution is that all tests must be passing. Aider can be configured with the command to run tests for a repo, and will automatically attempt to fix any test failures.

A user working on a python project might configure testing by launching aider like this:

aider --test-cmd pytest

For the benchmark, aider is configured with a test command that will run the tests that already exist in each problem’s repository. SWE Bench problems are based on repositories from large open source projects with extensive existing test suites. This means that testing will fail if aider has broken any of these pre-existing tests or if any new tests that it created aren’t passing.

As with editing and linting, aider reports a testing outcome that indicates if it completed with any outstanding failing tests. The benchmark harness uses this status when deciding if aider has produced a plausible solution.

To be clear, aider cannot run or even see the held out “acceptance tests” that are used to judge if a proposed solution correctly resolves the problem. Those tests are only run outside of aider and the benchmark harness, to compute the final benchmark statistics.

Finding a plausible solution

Each time aider executes, it reports the outcome of the editing, linting, and testing steps. Each of these steps may complete successfully or return a status that indicates that there were outstanding problems that remain unresolved.

The benchmark harness uses these outcomes to determine if aider has produced a plausible solution to the current SWE Bench task. A plausible solution is one where aider returns saying that it edited the repo with no outstanding edit, lint, or test errors. In this case, aider’s changes are recorded as the SWE Bench model_patch to be evaluated later with the acceptance tests.

If the solution is not plausible, another instance of aider is launched again from scratch on the same problem. The harness alternates launching aider with GPT-4o and Opus to solve the problem, and gives each model three attempts – for a total of six attempts. As soon as a plausible solution is found, it is accepted and the harness moves on to the next SWE Bench instance.

It’s worth noting that repositories may have lint or test errors present before aider even starts to edit them. Whether unresolved errors were caused by aider or were pre-existing, there will be instances where no plausible solution is found after six tries.

If all six attempts fail to produce a plausible solution, then the “best” solution available is selected as the model_patch. Which of the non-plausible solutions to use is determined by ignoring the testing outcome and prioritizing solutions in the following order:

Pick a solution where editing and linting were completed successfully.
Pick a solution where editing was at least partially successful and linting succeeded.
Pick a solution where editing was successful.
Pick a solution where editing was at least partially successful.

Computing the benchmark score

The benchmark harness produced a plausible solution for each of the 300 SWE Bench Lite instances and saved it as the model_patch.

A separate evaluation script was used to test each of these solutions with the full test suite, including the held out acceptance tests. For this final acceptance testing, any edits that aider made to tests are discarded. This ensures that the correct, unmodified test suite is used for acceptance testing. The evaluation script compares the test results with results from testing the “gold” patch that was developed by a human to correctly solve the issue. If they match, the candidate solution has correctly resolved the issue.

These acceptance tests are only ever run outside of aider and the benchmark harness, and only to compute the number of correctly resolved instances. They are never run, used, or even visible during aider’s attempts to solve the problems.

Aider correctly resolved 79 out of 300 SWE Bench Lite instances, or 26.3%.

Acknowledgments

Much thanks to the team behind the SWE Bench family of AI coding benchmarks. Also thanks to Albert Örwall who has dockerized the SWE Bench evaluation scripts making it faster, easier, and more reliable to run the acceptance tests.

References

All of aider’s results reported here are pass@1 results, obtained without using the SWE Bench hints_text.

The “aider agent” internally makes multiple “attempts” at solving the problem, but it picks and returns one single candidate solution. Only that one candidate solution is evaluated with the acceptance tests and contributes to the benchmark score. Thus it is a pass@1 result.

This is contrast to a pass@N result for N>1, where N attempts are made and all N solutions are evaluated by the acceptance tests. If any of the N solution pass, that counts as a pass@N success.

Below are the references for the other pass@1 unhinted SWE-Bench results displayed in the graph at the beginning of this article.

Note, the graph was corrected on 5/30/24 as follows.

The graph now contains AutoCodeRover’s average pass@1 results. Previously it displayed pass@3 results, which are not comparable to the pass@1 results for aider being reported here. The AutoCodeRover GitHub page features pass@3 results without being clearly labeled.

The graph now contains the best OpenDevin results obtained without using the SWE Bench hints_text to provide hints to the agent. The previous graph contained their hinted result, which is not comparable to the unhinted aider results being reported here. OpenDevin reported hinted results without noting that hints were used.