An entity-only index with relations materialized at inference reaches 95.6% on SWE-bench Verified with zero pre-built edges.

Google released Flash / Shared entities / Earlier coverage

Gemini 3.6 Flash cuts output tokens by 17%

Linked by a graph relationship (Google released Flash); both cover Flash, SWE, Verified; earlier Flash coverage from 2026-07-21.

Shared entities / Same source domain / Shared topic / Earlier coverage / Downstream implication

New paper trains RL through the agent harness, not the model

Both cover SWE, Verified; reported by the same outlet (arxiv.org); overlapping topics (agent, code).

Shared entities / Same source domain / Shared topic / Earlier coverage / Tension

Verify agent patches by reconstructing the problem backwards, with the original issue hidden.

Both cover SWE, Verified; reported by the same outlet (arxiv.org); overlapping topics (agent, inference).

Google released Flash / Shared entities / Same source domain / Earlier coverage

Qwen3.8 Max wins the agentic index and burns 64 turns doing it

Linked by a graph relationship (Google released Flash); both cover Flash, Verified; reported by the same outlet (arxiv.org).

Shared entities / Same source domain / Shared topic / Earlier coverage

9B model jumps to 56.4% on SWE-bench with 6K examples

Both cover SWE, Verified; reported by the same outlet (arxiv.org); overlapping topics (agent, base, code).

Shared entities / Same source domain / Earlier coverage

PMCoder resolves 25 more SWE-bench Verified cases

Both cover Flash, SWE, Verified; reported by the same outlet (arxiv.org); earlier Flash coverage from 2026-08-10.

Shared entities / Same source domain / Shared topic / Earlier coverage

Envs-FORGE lifts SWE-bench Verified score to 77.1%

Both cover SWE, Verified; reported by the same outlet (arxiv.org); overlapping topics (base, code).

Coding agents lose 7.7 points when you edit files mid-task

Both cover SWE, Verified; reported by the same outlet (arxiv.org); overlapping topics (agent, code).