diff --git a/tradingagents/agents/managers/research_manager.py b/tradingagents/agents/managers/research_manager.py
index c537fa2f..567394ac 100644
--- a/tradingagents/agents/managers/research_manager.py
+++ b/tradingagents/agents/managers/research_manager.py
@@ -9,6 +9,10 @@ def create_research_manager(llm, memory):
         sentiment_report = state["sentiment_report"]
         news_report = state["news_report"]
         fundamentals_report = state["fundamentals_report"]
+        
+        # Get LoRA-scored news sentiment
+        news_net_sentiment_score = state.get("news_net_sentiment_score", 0.0)
+        news_net_sentiment_label = state.get("news_net_sentiment_label", "Neutral")
 
         investment_debate_state = state["investment_debate_state"]
 
@@ -33,6 +37,9 @@ Take into account your past mistakes on similar situations. Use these insights t
 Here are your past reflections on mistakes:
 \"{past_memory_str}\"
 
+Additional Context:
+News net sentiment (LoRA-scored): {news_net_sentiment_label} (score: {news_net_sentiment_score:.3f})
+
 Here is the debate:
 Debate History:
 {history}"""
diff --git a/tradingagents/agents/researchers/bear_researcher.py b/tradingagents/agents/researchers/bear_researcher.py
index 6634490a..09a7c71e 100644
--- a/tradingagents/agents/researchers/bear_researcher.py
+++ b/tradingagents/agents/researchers/bear_researcher.py
@@ -14,6 +14,10 @@ def create_bear_researcher(llm, memory):
         sentiment_report = state["sentiment_report"]
         news_report = state["news_report"]
         fundamentals_report = state["fundamentals_report"]
+        
+        # Get LoRA-scored news sentiment
+        news_net_sentiment_score = state.get("news_net_sentiment_score", 0.0)
+        news_net_sentiment_label = state.get("news_net_sentiment_label", "Neutral")
 
         curr_situation = f"{market_research_report}\n\n{sentiment_report}\n\n{news_report}\n\n{fundamentals_report}"
         past_memories = memory.get_memories(curr_situation, n_matches=2)
@@ -37,6 +41,7 @@ Resources available:
 Market research report: {market_research_report}
 Social media sentiment report: {sentiment_report}
 Latest world affairs news: {news_report}
+News net sentiment (LoRA-scored): {news_net_sentiment_label} (score: {news_net_sentiment_score:.3f})
 Company fundamentals report: {fundamentals_report}
 Conversation history of the debate: {history}
 Last bull argument: {current_response}
diff --git a/tradingagents/agents/researchers/bull_researcher.py b/tradingagents/agents/researchers/bull_researcher.py
index b03ef755..30de3df2 100644
--- a/tradingagents/agents/researchers/bull_researcher.py
+++ b/tradingagents/agents/researchers/bull_researcher.py
@@ -14,6 +14,10 @@ def create_bull_researcher(llm, memory):
         sentiment_report = state["sentiment_report"]
         news_report = state["news_report"]
         fundamentals_report = state["fundamentals_report"]
+        
+        # Get LoRA-scored news sentiment
+        news_net_sentiment_score = state.get("news_net_sentiment_score", 0.0)
+        news_net_sentiment_label = state.get("news_net_sentiment_label", "Neutral")
 
         curr_situation = f"{market_research_report}\n\n{sentiment_report}\n\n{news_report}\n\n{fundamentals_report}"
         past_memories = memory.get_memories(curr_situation, n_matches=2)
@@ -35,6 +39,7 @@ Resources available:
 Market research report: {market_research_report}
 Social media sentiment report: {sentiment_report}
 Latest world affairs news: {news_report}
+News net sentiment (LoRA-scored): {news_net_sentiment_label} (score: {news_net_sentiment_score:.3f})
 Company fundamentals report: {fundamentals_report}
 Conversation history of the debate: {history}
 Last bear argument: {current_response}
diff --git a/tradingagents/agents/utils/agent_states.py b/tradingagents/agents/utils/agent_states.py
index 3a859ea1..a4b289d4 100644
--- a/tradingagents/agents/utils/agent_states.py
+++ b/tradingagents/agents/utils/agent_states.py
@@ -59,6 +59,16 @@ class AgentState(MessagesState):
     news_report: Annotated[
         str, "Report from the News Researcher of current world affairs"
     ]
+    # FinLLama News additions
+    news_items_scored: Annotated[
+        list, "Per-item news sentiment with confidence, relevance, weight"
+    ]
+    news_net_sentiment_score: Annotated[
+        float, "Weighted net sentiment score over fetched news items"
+    ]
+    news_net_sentiment_label: Annotated[
+        str, "Label for net sentiment: Positive/Neutral/Negative"
+    ]
     fundamentals_report: Annotated[str, "Report from the Fundamentals Researcher"]
 
     # researcher team discussion step
diff --git a/tradingagents/dataflows/news_parsers.py b/tradingagents/dataflows/news_parsers.py
new file mode 100644
index 00000000..8735d1ec
--- /dev/null
+++ b/tradingagents/dataflows/news_parsers.py
@@ -0,0 +1,148 @@
+import re
+from typing import List, Dict, Any
+
+
+def _extract_urls(text: str) -> List[str]:
+    url_pattern = re.compile(r"https?://[^\s)]+")
+    return url_pattern.findall(text or "")
+
+
+def _strip_md(s: str) -> str:
+    if not s:
+        return s
+    # Remove simple markdown bold/italics markers
+    return re.sub(r"[*_`]+", "", s).strip()
+
+
+def parse_global_news(raw_text: str) -> List[Dict[str, Any]]:
+    """
+    Parses global news text produced by get_global_news_openai into a list of items.
+    Expected patterns include enumerated bold headings like:
+      1. **October 25, 2025: "Headline"**
+         - Trading Relevance: ...
+         (source links)
+    Returns a list of dicts with keys: date, headline, relevance, sources, raw.
+    """
+    if not raw_text or not isinstance(raw_text, str):
+        return []
+
+    items: List[Dict[str, Any]] = []
+
+    # Find each enumerated bold heading and take content until next heading
+    header_iter = list(
+        re.finditer(r"(?m)^\s*\d+\.\s+\*\*(.+?)\*\*\s*$", raw_text)
+    )
+    if not header_iter:
+        # Fallback: try to split by lines that start with a date-like pattern in bold
+        header_iter = list(
+            re.finditer(r"(?m)^\s*\*\*([A-Za-z]+\s+\d{1,2},\s+\d{4}.*)\*\*\s*$", raw_text)
+        )
+
+    boundaries = []
+    for m in header_iter:
+        boundaries.append((m.start(), m.end(), m.group(1)))
+    # Add sentinel end
+    text_len = len(raw_text)
+    for i, (s, e, header_text) in enumerate(boundaries):
+        next_start = boundaries[i + 1][0] if i + 1 < len(boundaries) else text_len
+        block = raw_text[e:next_start].strip()
+
+        header = header_text.strip()
+        # Extract date and headline from header
+        date_match = re.search(r"([A-Za-z]+\s+\d{1,2},\s+\d{4})", header)
+        quoted_headline = re.search(r"\"([^\"]+)\"", header)
+        headline_after_colon = None
+        if ":" in header:
+            parts = header.split(":", 1)
+            headline_after_colon = parts[1].strip()
+            # Remove surrounding quotes if present
+            headline_after_colon = headline_after_colon.strip("\"“”")
+
+        date_str = date_match.group(1) if date_match else None
+        headline = (
+            quoted_headline.group(1)
+            if quoted_headline
+            else (headline_after_colon or _strip_md(header))
+        )
+
+        # Extract trading relevance line(s)
+        rel_match = re.search(
+            r"(?i)Trading\s+Relevance:\s*(.+)", block
+        )
+        relevance = rel_match.group(1).strip() if rel_match else ""
+
+        sources = _extract_urls(block + " " + header)
+
+        items.append(
+            {
+                "date": date_str,
+                "headline": headline,
+                "relevance": relevance,
+                "sources": list(dict.fromkeys(sources)),
+                "raw": header + "\n" + block,
+            }
+        )
+    return items
+
+
+def parse_stock_news(raw_text: str) -> List[Dict[str, Any]]:
+    """
+    Parses company-specific news text from get_stock_news_openai into a list of items.
+    Expected patterns include bold enumerated sections like:
+      **1. Topic**
+      Description ... (url)
+    Returns a list of dicts with keys: title, summary, sources, raw.
+    """
+    if not raw_text or not isinstance(raw_text, str):
+        return []
+
+    items: List[Dict[str, Any]] = []
+
+    # Find headings like **1. Something** or **1. Something** on its own line
+    header_iter = list(
+        re.finditer(r"(?m)^\s*\*\*\s*\d+\.\s*(.+?)\s*\*\*\s*$", raw_text)
+    )
+
+    if not header_iter:
+        # Fallback: split by numbered lines even without bold
+        header_iter = list(
+            re.finditer(r"(?m)^\s*\d+\.\s+(.+?)\s*$", raw_text)
+        )
+
+    if header_iter:
+        boundaries = []
+        for m in header_iter:
+            boundaries.append((m.start(), m.end(), m.group(1)))
+        text_len = len(raw_text)
+        for i, (s, e, header_text) in enumerate(boundaries):
+            next_start = boundaries[i + 1][0] if i + 1 < len(boundaries) else text_len
+            block = raw_text[e:next_start].strip()
+            title = _strip_md(header_text)
+            sources = _extract_urls(block + " " + title)
+            summary = block.strip()
+            items.append(
+                {
+                    "title": title,
+                    "summary": summary,
+                    "sources": list(dict.fromkeys(sources)),
+                    "raw": f"{title}\n{summary}",
+                }
+            )
+    else:
+        # Last resort: try to split paragraphs; each paragraph with a URL is an item
+        paragraphs = [p.strip() for p in re.split(r"\n\s*\n", raw_text) if p.strip()]
+        for p in paragraphs:
+            urls = _extract_urls(p)
+            if urls or len(p) > 120:
+                items.append(
+                    {
+                        "title": p.split("\n", 1)[0][:80],
+                        "summary": p,
+                        "sources": list(dict.fromkeys(urls)),
+                        "raw": p,
+                    }
+                )
+
+    return items
+
+
diff --git a/tradingagents/graph/setup.py b/tradingagents/graph/setup.py
index accf083c..b7c174cc 100644
--- a/tradingagents/graph/setup.py
+++ b/tradingagents/graph/setup.py
@@ -76,7 +76,8 @@ class GraphSetup:
 
         if "news" in selected_analysts:
             analyst_nodes["news"] = create_news_analyst(
-                self.sentiment_llm
+                # self.sentiment_llm
+                self.quick_thinking_llm
             )
             delete_nodes["news"] = create_msg_delete()
             tool_nodes["news"] = self.tool_nodes["news"]
diff --git a/tradingagents/graph/trading_graph.py b/tradingagents/graph/trading_graph.py
index c2a2960d..34c07bd8 100644
--- a/tradingagents/graph/trading_graph.py
+++ b/tradingagents/graph/trading_graph.py
@@ -103,7 +103,7 @@ class TradingAgentsGraph:
             try:
                 self.sentiment_llm = DAPTLlamaChatModel(
                     dapt_adapter_path=dapt_path,
-                    max_new_tokens=512,
+                    max_new_tokens=1024,  # Increased for longer reports
                     temperature=0.7,
                 )
             except Exception as e:
@@ -238,6 +238,10 @@ class TradingAgentsGraph:
             "market_report": final_state["market_report"],
             "sentiment_report": final_state["sentiment_report"],
             "news_report": final_state["news_report"],
+            # Persist FinLLama News fields if present
+            "news_items_scored": final_state.get("news_items_scored", []),
+            "news_net_sentiment_score": final_state.get("news_net_sentiment_score"),
+            "news_net_sentiment_label": final_state.get("news_net_sentiment_label"),
             "fundamentals_report": final_state["fundamentals_report"],
             "investment_debate_state": {
                 "bull_history": final_state["investment_debate_state"]["bull_history"],