Software & Code
Research on Software & Code: explanations, evidence, limitations and connections between papers.
31 results
A single repair plan helps coding agents resolve more tasks
PAUSE gives coding agents one plan before a code change; tests found higher repair resolution in two task suites.
A fixed-state test shows what helps coding agents repair code
A fixed-state coding repair study finds specifications help more than test feedback on function-level failures.
REM times and trims memory for coding agents
REM manages coding agents’ past experience; across three runs, resolution was higher than with no memory, but not significantly so.
A controller-side check can stop commands based on stale state
AIS and XAIR check whether industrial commands remain valid at release; software-controller tests show the value of checking where state is owned.
Future-frame clues help neural codecs compress video
GFSC adds limited future-frame context to video codecs, reducing bitrate at comparable picture quality in tested models and datasets.
Inherited software work exposes a gap in coding agents
SWE-Residual finds lower coding-agent success on inherited work than matched clean starts across repository and machine tasks.
Verifier acceptance can miss weak generated contracts
In the study, only 42.3% of accepted contracts cleared checks for vacuity, test cases and reference agreement.
PACE-Bench checks whether coding agents deliver full requests
PACE-Bench evaluates full PRD delivery in existing repositories; the top tested configuration completed 6 of 43 cases.
Inspect behavior and provenance before reusing a coding capability
A coding team might call a specialist agent that helped before or copy a skill file from another repository. Reuse can spare repeated work, yet a prior passing solution can also carry a way to game a test. The studies suggest a practical first step: check the capability’s behavior in the current task and trace its source. Evidence from an agent library, coder-to-coder imitation and GitHub skill copies supports that judgment in different tested settings.
VulnSAGE checks the code evidence behind security warnings
On 187 labeled reports, VulnSAGE rejected 89 of 97 false alarms and retained 85 of 90 real defects.
Match each software test to the failure it must expose
A generated regression test can pass on faulty code if it adopts the faulty output as its expected answer; fixing that bug may then make the test fail. A test can also reject a valid implementation or miss a concurrency bug if the revealing thread order never runs. Read each verdict against the comparison that produced it: which failure, implementation, and execution could the test distinguish? These three studies support claims about their injected faults, labeled implementations, and controlled schedules, not software reliability in general.
Place review where an uncertain signal would change the dataset
A brain scan can resemble another scan without proving that both belong to one person. A math problem can resemble a benchmark question without being copied from it. In either case, acting on resemblance changes the dataset. Across identity auditing, math-data cleaning, and multilingual example construction, the useful question is where an uncertain signal meets a decision. The studies place review at different points in those workflows and measure different outcomes, so their results support a placement principle rather than one shared estimate of review’s benefit.
Whole-file context helps coding agents resolve more issues
A file-selection method improved test-checked issue resolution over retrieval and compression in a 300-issue Python evaluation.
Training helps desktop agents coordinate clicks and code
OS-Helix trains desktop agents to coordinate clicks and code, raising task success on 361 Ubuntu desktop tasks.
Rondo tests compiler bugs that depend on earlier calls
Rondo tests ordered model calls to expose compiler bugs that depend on retained state across four deep learning frameworks.
A code-history check flags overlap in bug-fixing tests
BenchChecker pairs model text with repository history to detect possible training overlap in bug-fixing evaluations.
Blanc proves properties of deployed Ethereum bytecode
Blanc checks proofs against deployed Ethereum bytecode and proves contract properties under stated assumptions.
Reward hacking can spread from one coding agent to others
Tests show coding agents can pass evaluator-bypassing code onward, while review reduces hacks but can reject valid solutions.
Vulnerability fixes must also keep projects working
VulnBump tests whether coding agents remove dependency vulnerabilities while keeping real projects buildable and passing their tests.
Match software automation claims to the checks that produced them
A tool can trigger a concurrency bug, produce a patch that passes issue tests, or name a bug in a code review. Each sounds like success, but each workflow sees different evidence and uses a different check. Read an automation result by asking what behavior the study reached, what counted as success, and which task conditions limited that check. These studies support conclusions about their tested workflows; their outcomes answer different questions.
Reins steers thread schedules toward concurrency bugs
Reins guides thread scheduling with agent-derived constraints and triggers more injected bugs than the strongest baseline in one test setup.
A bug-based test finds limits in automated code review
CRBugBench tests whether review agents find confirmed bugs, revealing limits on context-heavy repository tasks.
A framework shows where AI infrastructure changes slow merges
A matched repository study finds infrastructure co-changes slow merges in one cohort but not in another.
Generated tests often fail across valid implementations
TestPrism shows why single-reference test checks can overstate coding-agent reliability across valid and invalid implementations.