SWE-Trail: History-Guided Trajectory Collection for Coding Agent Training
READ THE ORIGINAL SOURCEOpen original sourcehuggingface.co · Opens in a new tabSWE-Trail is a trajectory collector for coding-agent post-training. It uses earlier trajectories to decide where additional collection effort should go, while keeping that history out of the agent's context to avoid train-test mismatch. During a rollout, history identifies a promising state, where the agent proposes candidate actions that SWE-Trail briefly executes from the same checkpoint to choose the continuing route. Training a 124B student on these trajectories improved mean resolution over independent sampling by 2.33 points on SWE-bench Multilingual and 1.60 points on SWE-bench Pro.
One primary source; other links are context, not independent corroboration.
- Primary sourceHugging Face Daily Papers
- Discovery linkHugging Face
- Linked in sourceView arXiv page
- Linked in sourceView PDF
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Open full article view →- Original paper+45
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
SWE-Trail is a trajectory collector for coding-agent post-training. It uses earlier trajectories to decide where additional collection effort should go, while keeping that history out of the agent's context to avoid train-test mismatch. During a rollout, history identifies a promising state, where the agent proposes candidate actions that SWE-Trail briefly executes from the same checkpoint to choose the continuing route. Training a 124B student on these trajectories improved mean resolution over independent sampling by 2.33 points on SWE-bench Multilingual and 1.60 points on SWE-bench Pro.
A READER'S EVIDENCE CHECKHow to read this signalSix questions about claims, sources, risks, and open checks
This is a signal-level guide until claim-level evidence is available. “Not recorded” means the evidence has not been assessed here; it does not mean a result is false. How we check AI news →
- 01Who is making each claim?
The claimant and whether each claim was measured or asserted are not recorded in this public brief.
- 02What was actually tested?
Not recorded per claim. Check the source for sample size, tasks, model and benchmark versions, and reproducibility.
- 03Where are the sources?
Original link: huggingface.co. Brief basis: fetched source page. Primary page date and quote locations are not established in this panel.
- 04What could mislead?
Source claims not independently checked; results are reported for a single 124B student initialized from Ling-3.0-flash-base.
- 05What is at risk if wrong?
Not assessed per claim. Consider safety, privacy, cost, and adoption decisions before relying on the result.
- 06What would change the view?
No claim-specific check is recorded. An inspectable primary method or independent replication could change the assessment.
Key findings from the stored brief
- Using history to guide where collection effort is spent, rather than exposing it as agent advice, avoids train-test mismatch.
- SWE-Trail improves mean resolution over independent sampling by 2.33 pp on SWE-bench Multilingual and 1.60 pp on SWE-bench Pro.
- Without history, the same pipeline gives only +0.11 pp on SWE-bench Multilingual, so the gain comes from history-guided collection.
Source claims not independently checked; results are reported for a single 124B student initialized from Ling-3.0-flash-base.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
Related archive coverage
These relationships record a citation or topic overlap, not confirmation of the finding.
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
Post-training data for coding agents is increasingly built from sampled trajectories, and how those trajectories are collected shapes what models learn. The paper argues that reusing earlier rollouts to direct collection can improve trained models without leaking unavailable advice into the agent's context.
Limits and open questions
Source claims not independently checked; results are reported for a single 124B student initialized from Ling-3.0-flash-base.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.