Unverified50% confidenceFactExact time
On Qwen2.5-7B, the intervention that improved MGSM accuracy the most was precisely the one that sacrificed the fidelity of the response language.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTaking Qwen2-VL-2B as an example, the combined strategy integrating hard negative mining, false-negative mitigation and hard batch construction improves performance on MMEB by about 6.0% on average.64% similarUnverifiedSpatialOPSD internalizes privileged depth information into a single RGB forward pass through self-distillation, raising spatial reasoning accuracy from 45.9 to 51.9 on Qwen3.5-9B.63% similarUnverifiedMiMo V2.6 Pro is quite sensitive to quantization — the more aggressively it is compressed, the more coding errors occur and the worse the experience becomes.63% similarUnverifiedGQA and MQA save the most in long-context, large-batch, high-concurrency inference scenarios, but the compression benefit weakens in short-sequence or small-batch scenarios, and they carry the risk of reduced expressiveness.63% similarUnverifiedSequential Thinking MCP harms performance on Qwen3-8B, because native think reasoning blocks are already trained into the Qwen3-8B weights, and adding an additional external thinking tool leads to a double thinking loop, doubling latency and token consumption without improving analysis quality.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1134678API
curl https://kongchang.com/api/v1/knowledge/claims/1134678MCP
get_claim(id=1134678)