Translation pending. Opens in a new tab
Translation pending.
Thread groupedOpen source Opens in a new tab
River
Primary-source items first added here within the past 48 hours.
fix(smoke): fail release checks on unsuccessful shutdown
Problem and resulting behavior The Docker release smoke runner accepted any process exit during the SIGTERM grace period as successful teardown. A start command could report a
paper: failure-mode taxonomy of the 65-task corpus + contribution reframing
Summary Two coordinated improvements to the paper's framing, built entirely from the released 65-task corpus — no new experiments, no new model calls.
Translation pending. Opens in a new tab
Translation pending.
Thread groupedOpen source Opens in a new tabfeat(benchmark): add S24 dock-cotenant-crash trap (shared error boundary co-tenant unmount)
Translation pending.
Translation pending. Opens in a new tab
Translation pending.
Thread groupedOpen source Opens in a new tabdsh-failbook: A Failure Ledger Plugin for DeepSeek Harness Opens in a new tab
The dsh-failbook plugin automatically logs failures and intercepts retries to aid diagnosis. No fix confirmed yet.
UnclusteredOpen source Opens in a new tabdsh-eval-harness: YAML-Driven Regression Eval Gate | DSH Plugin
dsh-eval-harness: A Regression Eval Gate Plugin for DeepSeek Harness biboyang/dsh-eval-harness