Unverified50% confidenceEventExact time
In some diagnostic outputs Gemma tried to end a JSON string with a curly quote, causing it to be unable to stop and to repeat itself; the fixes included a per-field length cap, truncating at stray curly quotes, checking that the summary ends like a sentence, and reordering the fields to cause, summary, fix. After the fix it went from 2 of 5 usable to all 4 usable, taking 16–40 seconds.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/18/2026
First Seen
Valid until: 1/16/2027
Sources
Related Entities
Related Claims
UnverifiedOllama's default context silently truncates the beginning of long prompts, and that is exactly where the instructions are, causing early failures to look like Gemma ignored the specification.60% similarUnverifiedWhen a CommonMark parser encounters a fenced code block opening line with insufficient indentation, it treats it as a lazy continuation of the previous paragraph, causing the code block to collapse into an ordinary paragraph without reporting an error60% similarUnverifiedThe collapse in true-theorem recognition ability caused by SFT was observed across four repeated experiments with different random seeds and was reproduced on Gemma-3-4B.57% similarUnverifiedBehavior/format gaps (such as tone, JSON format, step consistency) are usually prompt problems, and most tickets that 'need fine-tuning' disappear after two weeks of rigorous prompt engineering.56% similarUnverifiedThe original article's scheme assigns weights to different failure categories: prompt injection failures weight 5, JSON formatting errors weight 3, multi-hop reasoning wrong answers weight 2, with the final score calculated on a weighted basis.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1061383API
curl https://kongchang.com/api/v1/knowledge/claims/1061383MCP
get_claim(id=1061383)