Update Automated MNLP evaluation report (2026-06-08)

EVAL_REPORT.md

@@ -2,23 +2,23 @@
 
 - **Model repo:** [`cs-552-2026-thinkinsidethebox/general_knowledge_model`](https://huggingface.co/cs-552-2026-thinkinsidethebox/general_knowledge_model)
 - **Owner(s):** group **thinkinsidethebox**
-- **Generated at:** 2026-06-06T17:33:41+00:00 (UTC)
+- **Generated at:** 2026-06-08T04:40:54+00:00 (UTC)
 - **Pipeline:** [mnlp-project-ci](https://github.com/eric11eca/mnlp-project-ci)
 
 _This PR is opened automatically by the course CI. It is **non-blocking** — you do not need to merge it. The next nightly run will refresh this file._
 
 ## Evaluated checkpoint
 
-- **Commit:** [`85504c2`](https://huggingface.co/cs-552-2026-thinkinsidethebox/general_knowledge_model/commit/85504c2e5486a253af83f73c379ccc2006ec6bb5)
-- **Message:** Upload tokenizer
-- **Committed:** 2026-06-04T11:02:30+00:00
+- **Commit:** [`a63624d`](https://huggingface.co/cs-552-2026-thinkinsidethebox/general_knowledge_model/commit/a63624d4ce801853890021d7e3284c3956f14b4b)
+- **Message:** Phase 3 DPO
+- **Committed:** 2026-06-06T19:50:56+00:00
 
 ## Summary
 
 | Benchmark | Accuracy | Status |
 |---|---:|---|
 | Math | — | not run |
-| Knowledge | 0.3500 | ok |
+| Knowledge | 0.3100 | ok |
 | Multilingual | — | not run |
 | Safety | — | not run |
 
@@ -30,30 +30,22 @@ _Prompts are intentionally omitted to avoid revealing benchmark contents. For mu
 
 **Correct** (1 shown)
 
-- **reference**: `G`
+- **reference**: `A`
 - **overall** (1/1 completions correct)
-- **extracted** (✓): `G`
+- **extracted** (✓): `A`
 - **completion**:
 
     ```text
-    <think>
-    
-    </think>
-    
-    \boxed{G}
+    \boxed{A}
     ```
 
 **Incorrect** (1 shown)
 
-- **reference**: `A`
+- **reference**: `C`
 - **overall** (0/1 completions correct)
-- **extracted** (✗): `G_f(\omega) = |H(\omega)|^2 / G_{\textphi}(\omega)`
+- **extracted** (✗): `D`
 - **completion**:
 
     ```text
-    <think>
-    
-    </think>
-    
-    \boxed{G_f(\omega) = |H(\omega)|^2 / G_{\textphi}(\omega)}
+    \boxed{D}
     ```