LLM judge (deepseek-v3) agrees with human evaluator on 91.6% of 200 sampled jailbreak responses

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• DeepSeek-R1 Llama 8b gains 0.16% accuracy on GSM8k with positive intervention (more reflections) at cost of ~2000 additional tokensfinding0.788
  Only model showing marginal benefit from increased reflection, at substantial token cost
• Five judge models agree 90-96% on multi-attempt detection and ESR direction for same responsesfinding0.775
  Validation that ESR findings are not artifacts of any particular judge model's evaluation methodology
• DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning (DeepSeekAI, 2025)concept0.770
  Paper introducing DeepSeek-R1 model and reporting self-reflection as aha moment
• All five judge models consistently rank Llama-3.3-70B as having substantially higher ESR rates than other modelsfinding0.762
  Cross-judge validation of the primary ESR finding across OpenAI, Alibaba, Anthropic, and Google judge models
• DeepSeek v3.2 increments bid from 10 to 850 over 49 sole-bidder roundsfinding0.761
  One DS-v3.2 trace shows extreme self-escalation, suggestive of treating own bid as competitor.
• Easy questions (acc > 80%) have average reflection rate of 25.8% for DeepSeek-R1 Llama 8b on GSM8kfinding0.755
  Baseline reflection rate for easy questions confirming difficulty-reflection correlation
• Zheng et al. 2023 - Judging LLM-as-a-judge with MT-Bench and Chatbot Arenaconcept0.749
  Source paper for the MT-Bench evaluation benchmark used to assess capabilities post-SOO fine-tuning
• DeepSeek v3.2 self-bidding rate 75.4%finding0.748
  DS-v3.2 has a high proportion of self-bidding rounds.